Intel 피드

Synchronisation and SMPTE timecode (time code)

unsloth/GLM-5.3-GGUF

Breaking Claude Code Opus 5 Auto Mode

Breaking Claude Code Opus 5 Auto Mode Anthropic are putting a great deal of faith in Claude Code's auto mode for protecting their coding agent users against prompt injection attacks. They recently made that the default and have made bold claims about its effectiveness. Johann Reh…

The Twelve-Factor App (2025)

incoai/GLM-5.3-Flash-DFlash2

Benchmarking Pocket-Scale Inference

CritICL: Inference-Time Weak-to-Strong Generalization from Small Language Model Failure Modes

Recent advances in inference-time scaling have significantly improved the reasoning performance of large language models (LLMs). However, these methods typically rely on repeated generation or external verification. To address this limitation, we introduce CritICL, a novel inference-time framework that improves reasoning while maintaining high efficiency. Our key insight is that LLM failure modes exhibit structured patterns across model scales within the same family. Instead of treating failures as undesirable outputs, CritICL leverages them as a source of guidance. Specifically, we utilize failure modes derived from weaker models and incorporate them into inference through critique-based in-context examples. We propose two variants: CritICL-dynamic, which adaptively predicts input-specific failure modes and retrieves critiques, and CritICL-static, which uses a global failure mode profile to provide stable guidance. Experimental results show that CritICL consistently outperforms standard in-context learning and achieves performance competitive with or superior to test-time scaling methods, while requiring significantly fewer generations and lower token cost. Code available at: https://github.com/umwyf/CRITICL

WikiSkill: Compiling Agent Experience into Persistent Knowledge for Skill Evolution

Agent skills package specialized knowledge and workflows into reusable resources that extend AI agent capabilities. Recent work automatically discovers such skills from agent experience, which enables agents to progressively adapt through interaction. However, the insights that guide skill development typically remain scattered across optimization histories, limiting their systematic reuse across iterations. We introduce WikiSkill, a framework that co-evolves agent skills with a persistent knowledge base (wiki). At a high level, WikiSkill separates raw execution experience, accumulated knowledge, and executable skills, while continuously consolidating experience into the wiki, which subsequent skill updates can build on. Across diverse benchmarks and models, WikiSkill consistently outperforms state-of-the-art skill-evolution methods and improves over no-skill baselines in most model-benchmark settings. We find that skill evolution complements model scaling: larger models generally benefit more from evolved skills, while smaller models with skills can outperform substantially larger models without them. We also find that evolved skills transfer effectively across models and model families, and skills evolved by other models can outperform self-evolved skills. Finally, our ablation studies confirm that persistent knowledge accumulation in the wiki is critical for effective skill evolution. These results demonstrate the benefits of systematically accumulating and refining agent experience for developing reusable and transferable skills.

TTPO: Test-Time Policy Optimization

Recent prominent post-training methods, such as Reinforcement Learning (RL) and On-Policy Self-Distillation (OPSD), have driven rapid progress in mathematical reasoning for large language models, yet their reliance on ground-truth labels precludes test-time training (TTT). Replacing ground truth with majority-vote pseudo-labels is a natural alternative, yet it is fragile: an incorrect vote corrupts the teacher and misleads every token. We observe that this failure mode is asymmetric: rollouts that disagree with the pseudo-label are typically wrong regardless of whether the vote itself is correct. Building on this observation, we propose Test-Time Policy Optimization (TTPO), an asymmetric objective that distills agreeing rollouts via OPSD and penalizes disagreeing rollouts with Grouped RL. Token-level selection further refines both branches: distillation down-weights already-converged positions, while RL penalizes only confident errors. Both updates remain well-grounded even under frequent pseudo-label errors, and majority-vote routing yields tighter self-supervision as the model improves. Without any labels, TTPO matches label-supervised OPSD on five competition-level benchmarks, raises Qwen3-1.7B from 38.0% to 45.2% in TTT, yields +25.2% to +36.4% without thinking, and shows strong cross-task generalization.

SWE-Prime: Fewer Trajectories, Better Performance

To improve large language models' ability to resolve real-world software issues, prior work has focused on constructing large-scale agent trajectory datasets and performing supervised fine-tuning (SFT) on successful trajectories. However, task success does not guarantee high-quality supervision: successful trajectories may still contain ineffective, redundant, or risky steps. Directly using such trajectories for SFT can introduce noisy supervision and encourage models to imitate undesirable problem-solving behaviors. Therefore, we propose SWE-Prime, a multi-granularity, two-stage SFT data selection method that progressively filters training data at the trajectory and segment levels. Specifically, the first stage performs trajectory-level screening based on process quality, result quality, and data representativeness, selecting a high-quality and representative subset of successful trajectories. The second stage performs segment-level selection by grouping consecutive steps into semantic segments and assessing each segment based on its contribution to the final solution, learnability, and potential risks. During SFT, all segments remain in the sequence to preserve context, while only selected segments contribute to the loss computation. Experiments on SWE-Bench Pro and SWE-Bench Verified show that training on the 10% trajectory subset selected by SWE-Prime outperforms training on the full resolved dataset, yielding relative performance gains of up to 12.2% and 24.2%, respectively.

From Static to Dynamic: Benchmarking Real-World Code Review with MCR-Bench

In real-world software development, code review typically involves iterative interactions between developers and reviewers to improve software quality, making the process costly and time-consuming. Although recent work explores large language models (LLMs) for automated code review, most approaches oversimplify code review into a single-round, static decision task, which fails to capture the multi-round interactive nature and the complex problem-solving processes inherent in realistic review scenarios. To bridge this gap, we introduce MCR-Bench, the first defect state-aware benchmark designed for realistic multi-round code review. MCR-Bench covers five commonly-used programming languages and consists of 2,269 real-world multi-round code review tasks, each of which is annotated with fine-grained defect information and cross-round state labels. Each task in MCR-Bench is equipped with fine-grained defect metadata (e.g., description, type, severity) alongside dynamic state annotations, capturing the complete evolutionary trajectory of a defect throughout the multi-round process. We obtain several findings through extensive experiments on MCR-Bench with mainstream LLMs. (1) Limited overall capability: experiments reveal that mainstream LLMs exhibit limited overall performance in defect detection and defect lifecycle state tracking, with performance degrading significantly as the number of interaction rounds increases; (2) Defect-sensitive performance: LLMs' performance varies substantially across different defect types and severity levels, with semantically complex or low-salience defects being significantly more likely to be missed; (3) Underlying Failure Mechanisms: our in-depth error analysis dissects the distinct drivers of false positives and false negatives, revealing critical weaknesses such as cross-round temporal misalignment and inadequate long-range memory.

RedEvoAgent: Automatic Red-Teaming Agent with Experience-Driven Skill Evolution

LLM-based agents are increasingly deployed in product-level execution harnesses, where jailbreaks can trigger harmful tool use and persistent state changes, creating greater risks than unsafe text generation alone. Existing automatic red-teaming methods often rely on fixed attacks, while recent agentic attackers coordinate multiple jailbreak tools and show stronger potential through trajectory-based retrieval. However, such retrieval can reuse misleading experiences due to retrieval bias and unclear tool credit, and full trajectories add context overhead while reducing interpretability. We propose RedEvoAgent, a black-box red-teaming agent that distills cross-case attack trajectories into a concise, human-readable attack skill. The attack skill adaptively evolves through tool-effectiveness profiling and Deciding-Tool Attribution for skill updates, and a validation ratchet that retains only updates improving validation performance. Experiments on multiple benchmarks, target models, and target execution harnesses show that RedEvoAgent outperforms fixed and agentic baselines, improves tool efficiency, and transfers across attacker models and target execution harnesses.

Mechanistic Reaction Prediction via Discrete Flow Matching on Graph-Structured Electron Occupation

Chemical reactions are fundamentally transformations in electron space, yet most machine learning approaches model them either through \textit{de novo} generation of product molecules or through heuristic graph edits that operate directly on molecular topology. We introduce MAELLE (\textbf{M}ech\textbf{A}nistic \textbf{E}dit f\textbf{L}ow-matching on e\textbf{L}ectron r\textbf{E}arrangements), which instead models reactions as discrete flow matching over electron occupation vectors. Concretely, we formulate the reactant-to-product mapping as a Continuous-time Markov Chain (CTMC) over the graph-structured integer-valued electron occupation space defined on all bonding, non-bonding, and hydrogen sites. To construct the intermediate edit trajectories, we generalize the discrete flow matching mixture path to discrete electron rearrangements using Optimal Transport, yielding a sequence of mechanistically interpretable edit moves without requiring elementary step annotations. MAELLE achieves competitive performance on the USPTO-480K benchmark compared with leading reaction prediction models. Beyond in-distribution accuracy, we evaluate robustness across two out-of-distribution settings - structural complexity and reaction type - and find that MAELLE maintains strong performance where existing methods degrade. Finally, because the learned flow operates over the full electron redistribution, MAELLE naturally recovers mechanistic trajectories that align with known chemistry and can predict side products of a reaction.

Stochastic Estimation of Transduced Language Models

Transduced language models (TLMs) compose a pretrained \emph{source} language model with a functional finite-state transducer to induce a language model over \emph{target} strings. Computing the probability of a target prefix under a TLM amounts to summing the source-model probabilities of all source strings that the transducer maps to target strings beginning with that prefix. This set can be exponentially large or infinite. Prior work uses a computational shortcut based on source prefix probabilities, then approximates the resulting sum with threshold-pruned beam summing. This produces a lower bound with unknown error. Instead, we resample source prefixes without replacement and reweight each selected prefix by the inverse of its inclusion probability. We show that applying this correction recursively gives an unbiased estimator of the target prefix probability and lets us estimate the mass lost by threshold pruning. Our beam-summing algorithm extends the retained source prefixes and samples which prefixes to keep, reducing their number as more probability mass is added to the running estimate. This can save computation and guarantees that the run halts with probability one. We evaluate the method on encyclopedic text and DNA against sequential Monte Carlo baselines that resample with replacement. It achieves a better compute--variance tradeoff on text and lower error at the same maximum number of particles on DNA. On a DNA-to-amino-acid transduction, it reduces runtime by several orders of magnitude relative to threshold-pruned beam summing and makes estimating prefix probabilities for long target strings feasible. Replacing threshold pruning with unbiased sampling in a published reading-time analysis substantially lowers the estimated corpus surprisal but leaves the published conclusions unchanged.

Persona-Execution Separation: An Architecture Pattern for Evolving LLM Agents under Execution Audit

Large language model (LLM) agents in governed organizations must let the persona (instructions, tone, self-presentation) evolve freely, while keeping execution (stateful, audited work) traceable. A single trust domain does not satisfy both cheaply. We present Persona-Execution Separation (PES): persona and execution reside in different trust domains, connected by a governed contract bridge. The persona is singly-homed and may drift; execution is faceless and audited. Status summaries may return; data bodies remain in the restrictive domain except a graded data-loss-prevention (DLP) exception; identity stays continuous. An approval matrix, DLP, and audit enforce the crossing. PES follows from three goals---free drift, execution traceability, and decoupling. Under LLM representational indistinguishability, any single-domain mechanism that meets all three must re-introduce typed change objects, an external gate, and a stable audit anchor: PES rebuilt at higher coupling cost. A development/pilot case in a regulated digital-employee platform records five decisions over one month, each with a rejected alternative. A mechanism check on the shipped implementation found no execution-side re-validation under persona perturbation (five model configurations) and no persona fingerprint on hard-asserted fields. A probe of a recovered pre-separation build found the governed execution path decoupled from the persona by omission, not by construction; a later wiring change could reverse that isolation, which PES makes an audited architectural rule. The pattern applies when multi-user deployment, execution audit, and expected persona churn hold jointly.

Beyond F1: Evaluating Coverage and Failure Recovery in AI Model Security Scanners

Static scanners are increasingly used to identify executable or otherwise unsafe content in machine- learning artifacts, yet conventional evaluation metrics characterize only cases where a scanner yields a usable security judgment. We evaluate ModelScan, ModelAudit, and Fickling using a controlled, artifact-backed benchmark on a synthetic corpus of 170 Pickle and PyTorch focused artifacts across 145 specimen families, 135 of which have binary security ground truth and 10 of which are intentionally malformed without labels. We explicitly distinguish non-N/A coverage, analysis completion, definitive security decisions, non-security findings, and unsupported outcomes. On labeled families, ModelAudit produced definitive security decisions for all 135 families (100%), Fickling for 110 (81.5%), and ModelScan for 67 (49.6%). Conditional on making a definitive judgment, ModelScan achieved 100% precision, recall, and F1. Fickling identified no unique true- positive families beyond those found by the combination of ModelAudit and ModelScan. Furthermore, for the 48 malicious families where ModelScan failed to complete its analysis, both ModelAudit and Fickling generated detections consistent with ground truth. These findings underscore the need to separate judgment accuracy from judgment availability, as well as incremental detection coverage from tool-level redundancy.

Boosting LLM Exploration via Weak-Model Guidance in RLVR

Reinforcement Learning with Verifiable Rewards (RLVR) significantly improves LLM reasoning but often causes a drop in policy entropy, leading to narrowed reasoning coverage and degraded pass@$k$ for large $k$. While existing methods mitigate this entropy collapse through algorithmic regularizations, cross-model non-parametric perturbation is also neglected. In this work, we propose a simple yet effective approach to preserve the generative diversity of LLMs during RLVR. Instead of relying solely on internal exploration, we force the target model to generate answers based on partial reasoning trajectories generated by a smaller, weaker language models. These unfamiliar prefixes effectively disrupt over-confidence and encourage the exploration of distinct reasoning paths. We empirically study the potential of outer prefixes, revealing the mechanism of the impact of distributional discrepancy to the exploration dynamics in RLVR training. Experiments across multiple mathematical benchmarks show that our method consistently outperforms vanilla RLVR. Notably, the performance gain becomes increasingly pronounced as $k$ scales up, demonstrating a substantial expansion of reasoning coverage. Furthermore, our approach efficiently mitigates entropy collapse without requiring additional SFT, intricate reward designs, or complex prompting.

Scaling Graph Neural Networks for Friend Recommendation: Multi-Hash User Embeddings and Temporal Neighbor Sampling

Friend recommendation is inherently graph-structured: the relevance of a potential connection depends on multi-hop social context rather than user attributes alone. However, deploying message-passing GNNs on a production-scale social graph with hundreds of millions of users and tens of billions of edges requires addressing numerous modeling and systems challenges. We present a scalable end-to-end GNN ranking system for production social graphs, focusing on two design choices that are critical in this setting: multi-hash ID embeddings and temporal neighbor sampling. Multi-hash embeddings are common for high-cardinality features, but industrial GNN systems typically either ignore trainable IDs or accept full embedding tables, exceeding 200 GB for our graph. We integrate multi-hash as the primary node representation, reducing the ID-embedding table size by more than 98 percent while preserving ranking quality. Temporal neighbor sampling is well understood in principle, but existing implementations scan full adjacency lists, which is a non-starter for users with tens of thousands of friends. We implement timestamp-sorted CSR storage with binary search, reducing the per-node temporal sampling cost from $O(deg(v) + k)$ to $O(\log(deg(v)) + k)$. Beyond these components, we show that this combination scales and yields measurable production impact. On a graph with 194M users and 28B edges, offline ablations isolate each design choice's contribution. In an online A/B test, our system increases friend additions from recommendations by 16 percent and unique friend adders by 11.5 percent over a strong production baseline. We release our framework for distributed training and inference on large temporal graphs.

Consolidating RLVR Capabilities Across Domains: A Deep Dive into Fusion Paradigms

Reinforcement learning with verifiable rewards (RLVR) improves specific capabilities of large language models, but covering multiple capabilities often involves training separate domain experts and subsequently consolidating them. We organize three fusion paradigms by the artefacts they reuse: Merge combines expert task vectors, Mix RL pools their datasets, and multi-teacher on-policy distillation (MOPD) uses both. Because they have largely been studied in isolation, how they compare and how to choose among them remain unclear. We compare all three using shared experts and data across model scales and a multi-domain benchmark suite. Although their average performance differs by at most 1.4 points, the gap reaches 8.6 points on a single benchmark, with domain-level variation tracking cross-domain relations visible in task-vector geometry. Training dynamics expose distinct constraints: Mix RL depends on domain mixture proportions, MOPD remains bounded by its teachers, and Merge compresses all expert updates into one. All three improve single-sample accuracy without measurable gains in solution coverage or losses in held-out capabilities. These results yield a practical guideline: use Merge when experts already exist and cheap fusion is paramount; Mix RL when training a unified model without experts, with domain proportions adjusted for cross-domain transfer; and MOPD when preserving domain-specific gains matters more than surpassing teachers or minimizing end-to-end cost.

CLAP: Cross-Embodiment Video World Models are Zero-Shot Physical Simulators

State-of-the-art action-conditioned video models are typically restricted to a single robot embodiment, preventing them from leveraging the vast corpus of heterogeneous video data that contains rich signals for learning generalizable physics. To bridge this gap, we introduce CLAP, a framework for cross-embodiment action-conditioned video generation capable of being trained on diverse, internet-scale videos across human and robotic agents. CLAP is grounded in the insight that universal physical laws govern spatiotemporal dynamics regardless of the actor. However, cross-embodiment learning is non-trivial because action representations vary sharply across robot platforms and are typically absent in human videos. CLAP addresses this fundamental challenge through the following core contributions. First, CLAP reconciles disparate action spaces using end-effector poses, language instructions, and latent actions. Second, to resolve their individual limitations, CLAP introduces a curriculum-based cross-embodiment learning recipe that first learns foundational physical priors across unlabeled video data using latent actions and subsequently grounds them in end-effector action spaces for zero-shot deployment to real-world tasks. Crucially, CLAP approaches or surpasses state-of-the-art single-embodiment video models in challenging environments like DROID. These performance advantages compound via few-shot adaptation to establish a novel paradigm for training single-embodiment video world models. Ultimately, CLAP delivers the most comprehensive suite of action-conditioned video world models to date - spanning diverse action-conditioning spaces (end-effector, language, and latent) and robot morphologies (including cross-embodiment, DROID, Bridge, bimanual YAM robots, and G1 humanoids). We open-source all code and models. Project Website at https://omni-clap.github.io .

How Language Models Organize and Structure Moral Knowledge

How do large language models (LLMs) organize moral knowledge? Models detect moral content broadly, but detection is a low bar. We ask whether they go further, distinguishing moral foundations from one another and organizing the relationships between them geometrically. We train six independent linear probes on open-weight language models, one per Moral Foundations Theory (MFT) category (care/harm, fair/cheat, lib/oppress, loy/betray, auth/subv, sanc/degrade), and examine how the resulting directions relate to each other in representation space. We find the directions neither collapse into a single moral detector nor isolate from one another. Rather, they span a near-maximal number of independent dimensions while sharing a positive common component. The shared component is the signature of integration, and it is moral-specific relative to a matched non-moral concept battery built identically (mean pairwise cosine 0.26 vs. 0.013). The geometry is consistent across architectures and scale and reaches its integration regime early in pre-training, well before probe accuracy saturates. The structure the model discovers shows no evidence of the individualizing/binding distinction predicted by Moral Foundations Theory (an underpowered test: only 20 candidate partitions exist) but rather reflects corpus statistics. Extending to moral dilemmas, each dilemma direction partially composes from its component foundations, at 2.7x a mismatched-pair baseline, while the majority of its variance encodes conflict-specific structure. The model represents moral tension itself, not a pre-resolved judgment.

Making Clinical Language Models Auditable: Concept-Guided Fine-Tuning for Robust Prediction

Clinical language models can achieve strong in-hospital accuracy yet fail under deployment shifts because they exploit note-specific artifacts (e.g., templates, separators, boilerplate) that do not reflect patient state. We propose CAST (Concept-guided Artifact Suppression Tuning), an SAE-based framework for auditable clinical text classification. CAST uses Sparse Autoencoders to expose sparse, human-auditable features from intermediate Transformer activations, labels SAE latents with an LLM-assisted interpretation pipeline and ICD-10 retrieval constraints, suppresses verified artifact latents via residual subtraction during fine-tuning, and provides post-hoc per-concept attributions for auditing model decisions. On MIMIC-IV discharge-note mortality prediction, CAST improves over its corresponding fine-tuned encoder baselines and remains competitive with strong LLM baselines, while producing a feature-level audit trail of the clinical concepts that support each prediction and the artifact concepts suppressed during training.

LeVJEPA: Efficient & Scalable Video Pretraining without the Heuristics

Video carries the temporal structure of the physical world, yet learning representations from it has remained computationally expensive: prevailing self-supervised methods either prevent representation collapse through architectural asymmetries, coupling an exponential-moving-average target encoder, a stop-gradient, and a capacity-limited predictor, or circumvent it by reconstructing masked content in pixel space. We introduce LeVJEPA, the first video encoder trained under LeJEPA's collapse-free objective, which dispenses with both. A single encoder is trained with an invariance loss over global and local views of a clip, regularized by SIGReg, which excludes collapse with a provable guarantee. The architecture reduces to an encoder and a projector, and the objective to a single hyperparameter. This formulation admits two properties. First, the cost of pretraining is governed by the number of tokens the encoder observes; uniform random token dropping renders this number small while simultaneously improving downstream accuracy. At matched epochs on identical data, LeVJEPA matches or surpasses V-JEPA 2 across ViT-S/B/L at 5.6 to 20.8x less pretraining compute, and at matched total FLOPs it exceeds the strongest video baseline by 7.6 points on ImageNet-1K while remaining competitive on motion-centric benchmarks. Second, since no asymmetry between branches is required, the encoder can be trained with block-causal attention at no measurable accuracy cost: temporal ordering becomes a property of the encoder itself. Against a compute-matched DINOv2 trained on frames of the same videos, LeVJEPA approaches the image-pretrained encoder on appearance-centric evaluation while nearly doubling its motion-centric accuracy. These results indicate that, once its computational overhead is removed, video becomes a viable and in several respects preferable substrate for general-purpose visual pretraining.

RATIO: A Benchmark for Retrieval Across Typed Ideation Operations in Scientific Literature

Retrieved scientific literature can serve as inspiration for both human and AI scientists. Inspiration can take different forms: prior work may directly suggest how to address a problem, or surface directions at different levels of abstraction - zooming out to a more general view or zooming in to a concrete realization. We introduce RATIO (Retrieval Across Typed Ideation Operations), a large-scale benchmark in which relevance is defined by three operations which we name ideation moves: Address retrieves potential approaches for stated problems, Broaden retrieves more general formulations, and Specify retrieves concrete instantiations. RATIO is constructed from millions of full-text scientific papers across CS literature via a general recipe that extends discourse-marker distant supervision - previously used only for classification - to corpus-scale retrieval, combined with extensive LLM and human vetting. Experiments show that operation-specific fine-tuning substantially boosts retrievers but leaves much room for further improvements. RATIO provides a scalable training and evaluation framework for retrieval components that support literature-grounded ideation, opening up new research avenues on scientific inspiration retrieval.

CorporateBench: Large-Scale Q&A Benchmarking with Temporal Knowledge Bases

LLMs are increasingly able to answer complex questions about enterprise-scale document collections. But evaluation is hard: companies don't want to share internal communications, and synthetic datasets have been overly simple. We present CorporateBench (CB), a human-validated multi-task Q&A benchmark whose scale approaches the conditions LLMs encounter in corporate communication networks, with evaluation corpora surpassing 230,000 documents. CB evaluates LLMs across two dimensions (information extraction and knowledge base querying) through four synthetically generated firms ranging from 12 to 10,000 employees. Each corpus is sampled from a temporally evolving knowledge base describing a consistent world, guaranteeing cross-document logical consistency even across hundreds of thousands of documents. We evaluate five LLMs on CB, revealing increasingly poor performance as input size approaches realistic scales. CB provides LLM developers a metric for corporate communication reasoning, filling a crucial gap in the benchmarking ecosystem.

Dad's Custom Atari Peripherals – By Jim Trageser

Token-Level Advertising

Generative AI is transforming how people access information, challenging traditional advertising mechanisms built around predefined slots. Towards generation-native advertising, we propose the Latent Advertiser Mixture Auction (LAMA), a token-level advertising mechanism that embeds advertiser influence directly into the generation process. Advertisers report local continuation values that induce advertiser-specific next-token policies, from which the platform decodes through a latent mixture while updating an allocation posterior. We show that LAMA satisfies Markov DSIC and IR, and achieves near-optimal KL-regularized welfare. We further develop a learning-based implementation that reconstructs the required reports online from learned local advantages and root values. Proof-of-concept experiments on real-world commercial-search query splits show that LAMA improves platform welfare and revenue while maintaining user-facing response quality, providing initial evidence for the feasibility of generation-native advertising.

D2C-Routing: Dimension-to-Composition Evidence Routing for Mixed-Origin AI-Generated Text Detection

AI-generated text detection is commonly framed as a binary document-level judgment about whether a text is human-written or machine-generated. This framing breaks down for mixed-origin writing, where content origin and expression origin may differ. We cast mixed-origin detection as dimension-to-composition source attribution, inferring content origin and expression origin before composing them into four collaboration types. We propose Dimension-to-Composition Routing (D2C-Routing), which routes content-side and expression-side evidence to supervised dimension heads before a learned gated composition layer predicts the final label. On MixD2C, a reconstructed split derived from the HART mixed-origin benchmark, our disclosed D2C-Routing-based detector system reaches 0.8603 four-way Avg TPR@1%FPR, 6.5 points above the same-split RACE-local rerun. Core ablations support the routing design, while error analysis shows that distinguishing AI-content/human-expression from fully AI-generated text remains the hardest boundary. Code is available at https://github.com/bystander563/d2c-routing-artifact.

Puro-2B: Poor Lab's Qwen2-1.5B Trained on RTX 5090 within $5090

Language model pretraining has become almost synonymous with prohibitive cost, placing it out of reach for much of the academic and open-source communities. Although strong open-source efforts already exist, including open-weight models and open-source training recipes, a cost-efficient, hardware-accessible, and open-source pretraining recipe has long been missing. Even at a small scale, training Llama-3.2-3B costs over \$1.5M, and reproducing SmolLM3-3B needs over \$700K. In this report, we present an open pretraining recipe designed to lower this barrier. Using this recipe, we train a collection of Puro-2B models from scratch on up to 1.4 trillion tokens with FP8 precision on consumer-grade RTX 5090 GPUs. The models in the collection differ in token budgets and selected recipe variants. Our best model is trained at a compute cost of less than \$6.9K and approaches Qwen2.5-1.5B performance under our evaluation protocol. This cost efficiency is enabled by a combination of approaches, including hardware selection, low-precision training, hyperball optimization, curriculum model averaging, and the data recipe. Beyond the recipe itself, we provide two additional results. First, across the Puro-2B collection, we derive a Puro Cost Scaling Law that relates training cost to average model performance; the fitted law suggests that about \$4.4K, less than \$5,090, is sufficient to reach the performance of Qwen2-1.5B. Second, as an end-to-end case study, we examine how pretraining data curricula shape downstream performance after post-training. Such controlled studies are enabled by having access to the full pretraining pipeline rather than model weights alone. We release the full training recipe for Puro-2B, including data, code, and model weights under Apache 2.0 at https://huggingface.co/collections/thu-pacman/puro-2b.

Successive Capacity Growth: Task-Complexity-Driven Width and Depth Expansion for Vision Transformer Encoders in JEPA World Models

Joint-Embedding Predictive Architectures (JEPAs) for world modeling typically employ fixed-size Vision Transformer encoders that are over-provisioned for simple tasks and under-provisioned for complex ones, with significant redundancy across attention heads. We propose Successive Capacity Growth (SCG), a method that starts from a minimal encoder (1 head, 2 layers, 283K parameters) and grows incrementally in width (adding attention heads for low-level semantic capacity) or depth (adding transformer blocks for higher-order semantic abstraction), driven by a task-agnostic test-and-verify mechanism that exploits function-preserving expansion to safely trial architectural changes and roll back if they do not improve prediction loss. The Sketched Isotropic Gaussian Regularizer (SIGReg) ensures that all learned semantic dimensions remain statistically independent and aligned with the predictive objective, preventing collapse even as the architecture grows. On a 60-dimensional multi-object dynamics task, SCG naturally triggers depth expansion, improving prediction loss by 20.3% over the fixed small baseline with 56 times greater parameter efficiency than scaling to the fixed large model; on a 2D navigation task, a single width expansion yields even an 23% improvement over the fixed large model. Across all three tested environments of increasing complexity, the adaptive encoder matches or exceeds the fixed small baseline, with zero false-positive expansions and bit-exact function preservation (ratio = 1.0, absolute difference = 0.0). The take-away is that JEPA world model encoders need not be pre-allocated at maximum capacity - they can grow successively as the task demands, achieving significant compute and data efficiency while maintaining representation quality.

KnockGS:interaction-Grounded Calibrationof Physical Gaussian Representations

Physics-integrated 3D Gaussian representations now allow reconstructed deformable objects to be simulated and rendered under explicit material models. Existing pipelines, however, assume that material parameters are known or manually specified, limiting their applicability when these parameters must be inferred from observed object dynamics. We propose KnockGS, an interaction-response PhysicalGS framework that estimates the elasticity and density scales of a 3D Gaussian object from its dynamics under a known applied force. Rather than treating physical simulation only as a forward process, we turn the force-induced response into a calibration signal: temporal response features are xtracted from the observed dynamics, the two material scales are estimated from those features, and the estimate is then frozen and written back into the same simulator so that it can be tested on an interaction it was never fitted to.We evaluate the framework on both parameter recovery and response-level fidelity. The estimated scales are compared against hidden ground truth, and the re-simulated object is measured against the target using 3D particle trajectories, response-curve statistics, and rendered-frame quality. Across five held-out material targets, our method recovers the scales substantially more accurately than response retrieval, global regression, or a fixed default material, and the frozen estimate remains predictive under interactions that differ in direction and in magnitude. Interaction response therefore carries enough information to calibrate material scales in physically grounded 3D Gaussian representations.Our study is a first step toward interactive PhysicalGS systems that calibrate a Gaussian asset whose rendered appearance and simulated response are consistent.

Sophistication in GenAI Use: Field Evidence from a Large Firm

We study how sophistication in generative AI (genAI) use varies among the back-office workforce of a large firm. Using proprietary data, we observe 713,564 employee prompts and their corresponding large language model responses from nearly 4,000 back-office employees across 15 functional areas over eight months in 2025. We document three main findings. First, senior employees exhibit more sophisticated genAI use, consistent with domain expertise complementing genAI capabilities. Second, sophistication varies considerably across functions and is highest in Strategy, Digital Innovation, and Project Management, three groups that share a focus on firmwide strategic initiatives and organizational change. Third, we observe neither improvements in sophistication over time nor lasting improvements following formal AI training, suggesting that sophisticated use can be difficult to change. Together, our study provides measures of and insights into sophisticated genAI use that managers can use to improve outcomes and that researchers can use in future research.

Your Voice Cloning System is Secretly a Voice Anonymizer

Speaker anonymization suppresses speaker-identifying attributes from speech while preserving linguistic content and quality. We propose repurposing XTTSv2, a multilingual voice cloning model trained on 27k hours of speech, for speaker anonymization without retraining. Our key insight is that XTTSv2's voice cloning capabilities preserve prosodic structure independently of speaker identity, enabling voice conversion by conditioning on a pseudo-speaker. We introduce an iterative refinement strategy that balances privacy and utility by maximizing a harmonic mean of speaker dissimilarity and intelligibility. Evaluated on seven European languages across CommonVoice and Multilingual LibriSpeech, our system achieves near-optimal privacy (EER $\approx$ 0.49), competitive intelligibility, and substantially better speech quality than dedicated anonymization baselines, while requiring no language-specific training. We release the code here: https://github.com/rm00cr/coqui-tts.

Reverse engineering my ADHD test

Understanding Evolution Strategies for LLM Reasoning: Broader Reasoning Coverage than GRPO

Evolution Strategies (ES) have recently emerged as a memory-efficient post-training paradigm for LLM reasoning. However, the optimization behavior of ES remains understudied, making it hard to define its advantage scope compared to mainstream post-training paradigms (e.g., Group Relative Policy Optimization (GRPO)). By systematically investigating ES dynamics and mechanisms, this paper first identifies a performance advantage of ES over GRPO, theoretically and empirically showing that ES can lead to broader reasoning coverage, thereby better exploiting the reasoning capabilities of pretrained LLMs. Theoretically, we show that verifier-projected Jensen-Shannon diversity across the ES population is helpful to higher Pass@K performances. Empirically, unlike GRPO, which exhibits entropy collapse, ES improves Pass@1 while attaining higher Pass@K than GRPO. We further develop a sequential GRPO-ES training strategy that combines GRPO's strength in Pass@1 with ES's gains in Pass@K. Second, we find that despite substantial whole-model parameter drift, the task-performance gains of ES are only contributed to a sparse subset of larger-magnitude updates. This functional sparsity suggests that large parameter movement need not imply widespread functional change, and held-out evaluations further show that it does not necessarily lead to catastrophic forgetting. Finally, we study how hyperparameter design affects the effectiveness of ES, demonstrating that ES requires a smaller population size in a larger LLM. These findings position ES as a distinct reasoning post-training paradigm rather than a less effective, memory-efficient alternative to GRPO.

INTENT-AS-A-TOOL Makes it Easy to Track Agentic Misalignment

As large language models (LLMs) are deployed as autonomous agents, safety failures increasingly involve consequential actions. We study agentic misalignment, where agents take harmful actions under goal conflicts and pressures. Using chain-of-thought (CoT) monitoring, we find that harmful execution is often preceded by intent signals in reasoning. However, post-hoc CoT labels are too coarse to show how intent changes during generation. We introduce INTENT-AS-A-TOOL, an approach that adds intent-targeted tools to give the model a dedicated channel for expressing commitment to a target behavior. The probability of calling an intent tool provides a judge-free, fine-grained signal of the model's tendency to pursue that behavior. Our results show that INTENT-AS-A-TOOL complements CoT monitoring, expands post-hoc CoT labels into dense trajectories, and identifies critical steps for online intervention. These findings suggest that action preferences are useful for tracking agentic misalignment during reasoning. Our code and data are accessible: https://github.com/RebeccaZhang22/intent-as-a-tool.

PAWBench: How Far Are We from Probabilistically Aligned World Modeling?

Recent video generation models are increasingly framed as world models. Many physical processes can unfold in more than one valid way. Therefore, a world model should reproduce not only a plausible trajectory, but also the distribution of possible behaviors under the same initial observation and action. We call this distribution-level requirement probabilistic alignment. However, existing evaluations largely assess individual-video plausibility and do not test whether repeated generations recover the correct distribution. This raises a central question: how far are current video generators from probabilistically aligned world modeling? To answer it, we formalize probabilistic alignment as a distributional criterion for world models and introduce PAWBench, a benchmark for evaluating video generators as stochastic samplers of world dynamics. We further introduce PAWEval, an outcome-level protocol that converts repeated video rollouts into empirical distributions over possible physical behaviors. Across 50 scenarios and eleven current systems, no model consistently matches the reference probabilities while recovering the range of valid behaviors. Having established this gap, we test whether language prompts, initial noise sampling, or model training can reshape the model's predictive distribution. We believe our work can serve as a foundation for future efforts to move towards probabilistically aligned world modeling.

Dwarf Fortress is getting the mother of all magic updates

Pair-Level Essay-Scale Republication and Reuse from Fragmented Historical Text Reuse: A Workflow Study on Eighteenth-Century Books and Newspapers

This paper addresses the recovery of essay-scale republication and reuse from fragmented text-reuse evidence, a setting whose central challenge is pair-level evidence consolidation and not fragment retrieval alone. The study focuses on a candidate set centered on essays by eighteenth-century Scottish philosopher David Hume, spanning books from ECCO (Eighteenth Century Collections Online) and historical newspapers. Because the input consists of fragmented reuse hits instead of clean document pairs, and positive coverage is inherently incomplete, we formulate the task as pair-level evidence consolidation into plausible transmission relations and compare three methodological families: a staged rule-based workflow, baselines (a decision tree and two direct LLM settings), and automated rule adaptation. On labeled ECCO--ECCO slices, pair-level feature aggregation alone already reaches 0.948 F1 on the main labeled slice, while the final workflow gives the strongest overall precision-recall trade-off among the tested rule stages. On the full ECCO--ECCO candidate universe, direct LLM baselines flag up to 14,886 pairs as reprints compared to 771 for the final workflow, behaving in this direct-prompt setup as high-recall candidate expanders rather than precision-controlled deployment classifiers. On ECCO--Newspaper, manual audit confirms all 176 predicted positives as genuine cases of republication or reuse, while issue duplication and source-side multiplicity reveal additional provenance structure. Under incomplete ground truth, auditable pair-level evidence consolidation provides a practical way to produce compact candidate spaces for historical inspection.

BTS-AgentBench: A Deterministic, Replayable Pipeline from Read-Only Telemetry Logs to Agent Benchmarks

Industrial sites contain large volumes of read-only telemetry, but few benchmarks specify how to compile these records into executable multi-turn agent tasks. We present a telemetry-to-episode construction method instantiated as BTS-AgentBench. The pipeline normalizes BTS metadata and raw histories into a read-only tool store, compiles static tasks with tool-derived gold answers and evidence, and lifts retained tasks into typed, bounded operator-facing episodes. The 532-row release adds clarification, goal revision, timestamp policy, quality-gated reporting, and evidence attribution while preserving the source computation and split. Coded contract preflight reports zero findings, and the construction-exclusion controller completes 0/532 rows. Two independent raw-to-episode builds match all 11 logical tool-store exports and reproduce the released 356/87/89 train/dev/test artifact exactly. Applying the shared construction path to XAI4HEAT produces 204 episodes; on its 41-row held-out test split, the controller completes 0 rows and the retained GPT-5.5 execution completes all 41. Code, artifacts, and replay reports are available at https://github.com/kjy7567/BTS-AgentBench.

A Finite Sample Analysis for Quantile Temporal Difference Learning in Distributional Reinforcement Learning

We establish a global finite-sample guarantee for synchronous quantile temporal-difference learning (QTD) in tabular distributional reinforcement learning. The proof separates two stability mechanisms. A global comparison argument, based on the order monotonicity of reward cumulative distribution functions and the $W_\infty$ contraction of the distributional Bellman operator, brings an arbitrarily initialized iterate into a local neighborhood. Inside that neighborhood, we linearize the QTD mean field. Its Jacobian is a nonsingular $M$-matrix, and the associated positive semigroup permits a variance-sensitive martingale analysis. For stepsizes $α_t=c(t+1)^{-a}$ with $a\in(1/2,1)$, the leading last-iterate fluctuation is of order $\widetilde O\bigl(T^{-a/2}/\sqrt{1-γ}\bigr)$ and has no polynomial dependence on the number of quantiles. The deterministic transient and the required burn-in can still depend on the smallest Bellman-target density, which is of order $m^{-1}$ in the worst case. The result therefore distinguishes sharply between the local stochastic fluctuation and the global sample complexity.

Verify Smarter, Evolve Further: Efficient Harness Evolution through Behavior-Aware Verification

Agent harnesses shape how language-model agents use instructions, tools, and runtime components, but adapting these harnesses requires costly verification. Existing propose-and-verify methods typically score every candidate on a fixed task set, wasting rollouts on unrelated behaviors and allowing aggregate scores to obscure specific regressions. We introduce HarnessLens, a budget-aware framework for automated harness evolution. HarnessLens jointly explores the task space and user-configurable components, derives candidate modifications from execution trajectories, and selectively verifies each candidate on behavior-relevant tasks using an attributable-evidence gate. Across three agent harnesses and four benchmarks, HarnessLens improves average held-out performance by 7.6-13.6% while consuming substantially less evaluation budget than competing baselines. These results demonstrate that behavior-aware verification with explicit attribution enables more reliable and sample-efficient harness evolution under constrained interaction budgets. Our code is available at https://github.com/jhxu5214/HarnessLens.

Difference-in-Differences on a Censored Rating Scale Can Manufacture an Effect: Evidence from a Pre-Registered LLM-Judge Audit

Audits of LLM judges certify a bias by contrasting matched conditions, and the strongest designs difference twice: a within-item contrast between two candidate responses, differenced again across a manipulated attribute, read off a bounded rating scale. We show that this endpoint is not identified on the scale that reports it. Each term of the double difference is censored by its own share, so the observed statistic confounds differential preference with differential attenuation: a severity shift common to both responses manufactures an interaction whenever the two censor it unequally, as unequal distances from the bounds make them, exactly where good stimuli place them. We exhibit the failure inside a pre-registered audit of a frozen pedagogy judge, sealed before the first of its 990 calls. The registered primary endpoint, the effect of a stated learner profile on the judge's scaffolding preference, is null: $+0.085$ points (95\% BCa $[-0.167, +0.353]$, $p = 0.684$). The audit's one nominally significant interaction, $+0.378$ ($p = 0.002$), is not identified as preference: a construction containing zero differential preference reproduces 79 to 85\% of it from the observed severity shift and the scale floor alone. We derive the mechanism in closed form and show that its contribution is measurable from an audit's own ratings.

QuantumBoostNet: A Hybrid Classical-Quantum Architecture for Enhanced Accuracy in Cardiac Ultrasound View Identification

Accurate identification of the correct view or angle in cardiac ultrasound (echocardiogram) is a critical component of cardiologic imaging. This step is essential for precise anatomical interpretation, reliable measurement, and the reduction of clinical errors. Although computer vision has advanced significantly, most state-of-the-art models perform well on standard benchmarks but often yield suboptimal results in specialized medical imaging tasks due to the high level of noise present in the data. QuantumBoostNet, a hybrid classical-quantum architecture, is introduced to address these challenges. This model integrates a classical backbone with two heads: one classical and one quantum, with the quantum head implemented as a parametrized 10-qubit quantum circuit. Training occurs in two stages, with an adaptive transition between heads governed by a mixing parameter that monitors loss dynamics. Extensive experiments indicate that, despite the limited number of qubits that can be simulated, QuantumBoostNet consistently outperforms state-of-the-art classical and hybrid classical-quantum models in cardiac ultrasound view identification, achieving a relative improvement over the best competitor. QuantumBoostNet also demonstrates superior performance on established image classification benchmarks and exhibits robustness to noise. These findings support the continued development of hybrid classical-quantum models for specialized medical imaging applications.

Formalization of the Solution to the Hopf Problem

LLMs Can Design Near-Optimal OR Algorithms

We ask whether large language models (LLMs) can design effective algorithms for well-specified operations research (OR) problems. We study inventory control, queueing network control, and assortment optimization. We evaluate two levels of LLM use: at level 1, the model receives one problem instance and returns a solution for that instance; at level 2, it receives only the problem class description and broad parameter ranges, and returns an algorithm that maps instance parameters to solutions. Human input is minimal: we give one untuned prompt that describes the problem, and the model has access to a Python sandbox tool with a fixed compute budget. The strongest model we test, gpt-5.6-sol, matches or outperforms the best existing method on almost all evaluated instances. This holds even at level 2, where the returned algorithm is fixed before seeing the evaluation instances. Performance also improves sharply across models released less than eight months apart, suggesting that this capability is moving quickly. Thus, for the well-specified operations problems we study, a single untuned LLM query can already produce algorithms competitive with specialized methods. These results suggest that frontier LLMs can be a serious empirical baseline for algorithm design in well-specified OR problems.

Recovering Expert Critic-Sourced Network Adjacency between Musical Artists from Acoustic Distributions: A Construct-Validity Approach

Music recommendation relies primarily on two signals: user-item interactions, which fail in the cold-start regime, and intrinsic musical content, available for any recording. We argue that a third, largely untapped signal is both richer and more principled: critical adjacency, the pairwise relation established when an expert critic explicitly links two artists in long-form prose. It encodes deliberate judgments about which artists belong together. Prior work established its internal validity, showing it recovers coherent, interpretable communities and can match collaborative filtering in user-satisfaction simulations, with no user data. What has been missing is external validation: whether this critic-sourced relation is grounded in the music itself versus sociological context. We test it against acoustic content, reframing the question as one of construct validity. Representing artists as empirical distributions over 80 low-level Essentia acoustic descriptors and modeling pairwise proximity via marginal optimal-transport (Wasserstein) distances, we evaluate how far critical adjacency is sonically recoverable under a cold-start, artist-disjoint split. Our ensemble recovers these edges at out-of-sample AUC of 0.767 (95% CI 0.761-0.775). Recoverability rises monotonically with critical consensus, reaching 0.865 on multi-source attested edges. Stratified evaluations align with sociological models of genre: tightly bounded, scene-based genres show higher recoverability than broad industry umbrella terms. Critical discourse is thus a rich source of information for recommendation, decomposing into a reproducible "sonic core" and a "sociological remainder" driven by narrative positioning, subcultural context, and canonical placement. The work offers both a scalable cold-start discovery mechanism and a sociologically grounded approach to MIR and MRS research.

TADP: Task-Aware Deformable Prediction for Single-Stage 3D Object Detection

Most single-stage 3D object detectors complete different tasks with the same extracted features. Nevertheless, it is impossible to project features into a common space that is adaptive for all the tasks. We present a novel task-aware deformable prediction (TADP) method for single-stage 3D object detection to solve this problem. Firstly, a triple feature refinement aggregation module is designed to extract three-level features adaptively. Additionally, we design the multi-scale feature aggregation block to fuse multi-scale features in a scale-aware manner. Finally, the prediction of each task is deformed with the designed plug-and-play task-aware deformation head. It can percept the emphasis and interaction of each task. We also designed three different deformation modules. The experimental results demonstrate that the proposed deformation head shows good results on other detection methods. The experimental results on the KITTI dataset demonstrate that the car mAP is 80.91%, surpassing many state-of-the-art methods on the KITTI benchmark.

BrailleBench: Investigating Multi-Criteria Braille Comprehension in Large Language Models

Although Large language models (LLMs) mediate access to knowledge and computational assistance, their capabilities should benefit vulnerable groups in the same way. However, it is unclear whether existing AI systems are inclusive enough for blind and deafblind users to access the same functionality through Braille, whose indicators, contractions, and digital representations introduce distinct requirements for model comprehension. To this end, we introduce BrailleBench, a benchmark for evaluating LLMs in Braille comprehension from different Criteria. BrailleBench aligns 5,570 instances from five datasets, including mathematics, commonsense, and multi-hop question answering across English and Braille Grades 1 and 2. Different configurations are designed to understand whether the systems can comprehend Braille-authored content, express answers in Braille, and complete end-to-end Braille interaction. To ensure the quality and prevent evaluation bias, the benchmark is built through a deterministic, expert-reviewed pipeline via a self-created Braille Toolkit without using any data instances generated by LLMs. We evaluate six representative LLMs from various aspects. The results reveal a persistent gap between print-English capability and Braille accessibility. Braille understanding and expression are asymmetric, where Grade 2 is especially fragile on the input side compared to Grade 1, and fully Braille requests further reduce performance. The experimental observations provide valuable guidance for the development of future Braille AI systems. All related resources in BrailleBench are publicly available for future research.

Naive Prompt Optimization: Rethinking the Need for Complex Prompt Search

Efficiently improving autonomous agents across diverse tasks is central to accelerating recursive self-improvement (RSI) in agentic AI, with prompt optimization emerging as a promising approach capable of delivering performance gains comparable to those achieved by fine-tuning model weights, while reducing computational costs in both optimization and serving. However, recent developments increasingly favor unnecessarily complex prompt optimizers. We introduce Naive Prompt Optimization (NPO), a lightweight single-lineage method that iteratively revises prompts using a teacher model with rollout feedback. NPO achieves comparable or better performance than GEPA with fewer rollouts, and its advantage increases with stronger teacher models, suggesting that stronger teacher reasoning can partially substitute for optimizer-side search complexity. In interactive games, NPO remains broadly competitive with GEPA, while GRPO performs better on some tasks less amenable to prompt optimization. We also show that NPO-optimized prompts elicit similar performance improvements when applied verbatim to other student models, especially across models within the same family. Overall, our preliminary results show that simple, linear prompt optimization can rival substantially more sophisticated and complex search procedures.

SCIT: Testing Causal Cache Carriers in Latent Chain-of-Thought Models

Latent chain-of-thought models move intermediate reasoning from emitted text into continuous states, improving compactness but hiding the causal object. We introduce SCIT, the Suffix Cache Interchange Test, a causal protocol that constructs exact source-recipient counterfactuals, patches declared cache segments, and identifies which transformer object carries the counterfactual computation. SCIT combines sufficiency tests with K/V component splits, hidden-state controls, semantic source controls, decoded validation, and matched corruption. On CODI-GPT2 and a Sim-CoT-style GPT-2 reproduction, counterfactual arithmetic transfers primarily through value-cache suffix trajectories rather than hidden states, keys, reusable answer slots, or single-token triggers. Complete sufficiency-and-necessity evidence for the late-value-suffix mechanism holds for the main CODI-GPT2 checkpoint; the Sim-CoT-style checkpoint shows the same sufficiency and decoded-control pattern but insufficient matched-corruption evidence for a necessity call. Beyond these local arithmetic cells, SCIT reveals carrier-regime shifts: arithmetic-like GPT-2/1B cells preserve latent-tail value/KV transfer, whereas competent 8B and repaired non-arithmetic cells route through prompt-prefix or full-cache K/V; boundary cells receive no mechanism call. SCIT therefore contributes a cache-level diagnostic, a checkpoint-specific GPT-2 arithmetic mechanism, and a competence-gated carrier map rather than a universal latent-tail claim.

What Makes Good Agentic Data? An ACE Lens on Data Generation for LLM Agents

LLM agents increasingly rely on generated interaction data to learn how to interact with external environments. Agentic data generation must maintain consistency among environments, tasks, interactions, and success signals while producing experience that is useful rather than merely abundant. Existing work spans many agent domains, but domain-centered organization and heterogeneous evaluation often obscure common generation mechanisms and conflate candidate construction with verification and selection. This work develops a two-level framework for the field. First, we represent agentic data as a common factorized object $(E,q,τ,v)$, comprising an environment specification, task signal, interaction realization, and optional verifier. We organize generation paradigms by their primary anchor and dependency structure. Second, we formulate generation as constrained distribution design through the Accuracy-Complexity-divErsity (ACE) lens. Accuracy establishes the feasible support of grounded and internally consistent data. Within this support, Complexity places learning mass relative to the capability of a declared learner and execution configuration, while divErsity controls coverage and redundancy of data. Using this framework, we explore how prior work verifies generated experience, constructs and calibrates difficulty, and expands behavioral coverage. The literature reveals a shift toward execution-grounded accuracy, learner-relative complexity, and diversity beyond surface variation or dataset size. We further discuss broader directions and emerging trends in agentic data generation through the ACE lens, including their implications for scaling, data sources, training regimes and adaptive learning. Overall, the central challenge is not simply to generate more data, but to continually allocate valid, informative, and non-redundant experience as agents and environments evolve.

Compositional Online Learning for Semantic Data Processing Systems

An LLM call in a semantic data processing system is expensive enough to dominate query cost, yet slow enough to hide a CPU-side learner's update behind its round-trip. In production, LLM compute accounts for $80-90\%$ of query cost, and each call costs $10^5-10^7\times$ a relational predicate. The latency window inverts a design constraint of classical adaptive query processing, where online learners had to stay lightweight to avoid dominating the predicates they optimize. At LLM latency, per-call gradient steps and per-batch threshold solves fit inside the round-trip. We develop compositional online learning at the LLM call boundary: a framework for combining online-learning components in semantic data processing systems. Each component makes execution-time decisions and refines its learned artifacts online. The design space spans two axes, decision granularity and learner update cadence, and the components share a single learning pattern that hides each trainer step inside the next LLM round-trip. A production case study in Cortex AISQL composes three components: a memoization layer, an online per-call filter-ordering learner, and an online per-batch cascade-routing learner. A conditional cost decomposition assigns each learning component to a distinct factor of per-row LLM cost. Under independence, the two learning components compose multiplicatively to an $11.4\times$ upper bound on a representative conjunction-filter workload. Self-selection at the cascade boundary, sample-budget shrinkage, and selectivity-estimation drift reduce it to a realistic figure near $8\times$.

Importance Scoring of Transformer Attention Heads in Learning Tabular Data

Computationally demanding and opaque deep learning models can be better understood and optimized by analyzing how they transform data. While deep transformers have been widely studied in computer vision and natural language processing, their application in tabular data remains relatively underexplored. This paper presents one of the first applications of an importance-scoring metric to interpret multi-head transformer models in learning from tabular data. Experiments conducted on 40 diverse tabular datasets demonstrate robustness to head drops based on the proposed head importance score. In 72.5\% of experimental examples, the model remains most resilient to performance drops when heads with the lowest importance scores are gradually removed. In contrast, removing the most important attention head first results in the greatest reduction in classification performance. A closer look at individual head importance scores across six attention layers reveals that important heads are scattered across layers, with no consistent layer-specific trends. In contrast to the image and language domains, the importance of individual attention heads varies considerably across tabular datasets with different schemas and feature spaces. The proposed importance score can improve efficiency and redundancy within transformer architectures. We make the source code for measuring the importance of individual attention heads publicly available.

A Point-of-Prescription Safety-Check System for Adverse Drug Reactions in Rural Bangladeshi Hospitals: A Feasibility Study

Adverse drug reactions (ADRs) are a major, largely preventable source of patient harm. In high-income settings, electronic health records store a patient's allergy history and warn prescribers when a contraindicated drug is ordered; in rural Bangladeshi public hospitals no such record exists for outgoing patients, a single physician may see on the order of one patient per minute, and a patient's history of severe reactions does not survive between visits. This paper proposes and outlines the evaluation of a lightweight, smartphone-based safety-check system for this setting. At registration a soft identifier (a phone number) is recorded; after the physician writes a prescription, its image is captured, the brand names are resolved to active ingredients using national drug references, and the ingredients are matched against the patient's recorded severe reaction history. The system is retrieval-based rather than predictive, and is silent by default, raising a flag only for high-risk matches a design grounded in the alert-fatigue literature. We frame the work as a feasibility study: we describe the proposed framework and an evaluation plan measuring workflow fit under high volume, usability, identity-resolution reliability, and retrospective detection of known reaction cases. We explicitly do not claim a clinical-outcome effect, which the low base rate of severe events places beyond a single-site feasibility study.

HALO: A Heterogeneity-Aware Language-Aligned IMU Foundation Model for Open-Set Human Activity Recognition

Human Activity Recognition (HAR) using inertial measurement units (IMUs) enables a wide range of applications, yet the field still lacks a unified model that can generalize across diverse subjects, devices, and activities. Training such a model is difficult due to two key challenges: sensing heterogeneity -- differences in sampling rates, channel configurations, and sensor placements -- and poor generalization to unseen activities and label vocabularies. We introduce HALO (Heterogeneity-Aware Language-aligned Open-set model), a domain-specific IMU foundation model that addresses both challenges through a two-stage training framework. Stage 1 pretrains the IMU encoder with heterogeneity-aware self-supervised learning, including adaptive-pooling tokenization, channel-independent feature extraction, and contextualized sensor conditioning that injects natural-language sensor descriptions into each channel embedding. Stage 2 aligns this IMU encoder with text embeddings via synonym-aware soft contrastive learning, enabling open-set recognition via cosine-similarity retrieval without per-dataset classifiers. Trained on 10 public HAR datasets and evaluated on 7 held-out datasets, HALO outperforms five state-of-the-art baselines on all 8 aggregate metrics, and still leads on 3 of 4 settings under baseline-matched inputs. Despite using only ~35M trainable parameters -- 10x fewer than the latest foundation model MOMENT (341.2M) -- HALO improves zero-shot open-set accuracy, measured over all 87 training labels, by 13.7 percentage points. On two further datasets with severe distribution shift, every model including HALO collapses zero-shot. A video demonstration of HALO's performance in real world is available at https://youtu.be/rooVKragtFU

STEP: State-Aware Task Estimation and Planning with Multi-Modal LLMs for Human-Robot Collaboration

Effective human-robot collaboration in industrial settings requires robots to understand human intentions and assist with task planning, reducing workload. Recent works have explored the use of Multi-modal Large Language Models (MM-LLMs) for task planning in such data-scarce scenarios, leveraging in-context learning to interpret user actions and generate long-horizon action plans in natural language. However, MM-LLMs inherently lack an understanding of system states and do not track state transitions, often leading to hallucinated actions that deviate from the intended goal. Additionally, generating action plans in natural language tends to limit the generated plans to a high level, introducing ambiguity in action execution. To address these limitations, we propose the State-aware Task Estimator and Planner (STEP), which prompts a MM-LLM to explicitly estimate the state of the system and predict the state transitions resulting from executed actions. By forecasting future states alongside actions, STEP ensures task-convergent planning while also providing additional assistance parameters necessary for executing the predicted actions. We evaluate STEP in a simulated environment using a robot assembly task. Our approach outperforms the state-of-the-art by 32.8% in action executability and 14.8% in final-state error.

BALMS: Benchmarking Agentic LLMs for Longitudinal Mental Health Sensing

Mental health assessment relies on episodic self-report scales, which convert subjective states such as stress into numerical scores but provide only sparse snapshots of wellbeing. Wearable devices offer longitudinal behavioral and physiological signals for continuous, low-burden monitoring. Recent LLM-driven personal-health agents enable natural language queries over wearable signals, but mainly handle short-term, retrieval-based lookups (e.g., highest step count over a week). They do not evaluate whether agents can reason over long-term signals to predict wellbeing scores paired with evidence-grounded rationales. To address this gap, we introduce BALMS, the first systematic benchmark of LLM-based agentic systems for longitudinal mental health sensing. BALMS spans 3 real-world longitudinal datasets, 2 task families (closed-form wellbeing-score prediction and rationale generation auto-graded by an LLM-as-Judge), 3 agentic paradigms evaluated across 5 open- and closed-source LLM backbones. We find that zero-shot agents rarely outperform a simple mean baseline, except with stronger backbones or compact, semantically meaningful features. Chain-of-thought prompting improves reasoning-oriented backbones, but does not guarantee temporal grounding or numerical correctness. Together with more analysis on efficiency and temporal scaling, BALMS highlights the need for longitudinal mental health agents that selectively retrieve history, ground temporal evidence, and reason over interpretable behavioral features.

PACE: A Unified Condense-and-Extract Paradigm for Fast VLM Inference

Vision-Language Models (VLMs) demonstrate exceptional visual reasoning capabilities, yet their inference costs escalate rapidly with the proliferation of visual tokens. Existing visual token pruning methods exhibit two fundamental limitations. First, most approaches operate exclusively post-vision encoder, leaving the substantial latency of the visual encoding phase unoptimized. Second, under strict token budgets, these methods often fail to jointly preserve holistic visual contexts and fine-grained details, leading to performance degradation. To address these bottlenecks, we propose PACE (Pixel-Adaptive Condense and Extract), a training-free inference framework that accelerates both the vision encoder and the Large Language Model (LLM) via a unified Condense-and-Extract paradigm. During the Condense stage, an Adaptive Pixel Compressor (APC) evaluates visual information density prior to encoding, adaptively downsampling redundant inputs, curtailing encoder computation while preserving global context and essential visual cues. In the Extract stage, a Dynamic Dual-Attention Extractor (DDAE) selectively retains visual tokens via a fusion of internal visual signals from the encoder and semantic signals from the LLM, safeguarding task-critical details. By integrating PACE into Qwen2.5-VL-7B, the model retains 93.8% of its original performance while utilizing only 10% of the visual tokens, yielding a 3.1x speedup in time to first token (TTFT). Our code is available at https://github.com/jjL357/PACE.

Profit based evaluation of machine learning for nitrogen recommendations in winter wheat

Nitrogen rates for winter wheat are set before the season, under unknown prices and weather. The standard UK advice does not respond to prices, yet recent price swings moved the most profitable rate by tens of kilograms per hectare. Machine learning is often proposed as the fix. However, it is usually judged on prediction accuracy, and accurate prediction does not by itself make the recommended rate more profitable. Our insight is to score nitrogen advice directly by the profit it forgoes on measured yield response curves. We build a test bench on 892 such curves from two long running UK experiments, and sweep the nitrogen to grain price ratio to cover all price scenarios. On this bench, machine learning fails as a predictor. No model recovers the best rate within farm tolerance, and the benchmark noise shows none can. At normal prices, every model also loses to the standard advice on profit. The gain sits elsewhere. A simple correction step applied after the model cuts profit losses by a quarter, while better models and extra features give no gain. The same frozen correction cuts losses by 43% at the second site without any retraining. A hybrid of standard advice plus a damped correction removes bias and trims rare large losses. The same price sweep also prices emission cuts, at a cost comparable to current carbon prices. Machine learning therefore pays as a profit scored correction to standard advice, not as its replacement.