# Research Papers — 2026-09 · bundle page 240 of 266 > Full-text markdown of papers #11951–12000 of 13298 published in 2026-09, newest first. AI-readable mirrors of the full 13298-paper 2026-09 archive (50 papers per file). Canonical page: https://aipod.dev/papers/archive/2026-09/page/240 [← Previous bundle](https://aipod.dev/papers/archive/2026-09/page/239.md) [Next bundle →](https://aipod.dev/papers/archive/2026-09/page/241.md) ### Temporal Self-Distillation: Learning Visual State Tracking in Videos Without Supervision Authors: Shravan Venkatraman, Wenshuai Zhao, Mohammad Hassan Vali, Arno Solin Published: 2026-09-03 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.04203v1 We introduce S$^3$T (Self-Supervised Self-Distillation over Time), which, to the best of our knowledge, is the first fully self-contained framework for continuous video state tracking. Our method treats temporal sampling density as privileged information, based on the hypothesis that a denser view of the same clip recovers the running state more accurately. This view serves as the teacher, while a sparse-view student with the same weights learns to match its next-token distribution. The model generates its own target, so training requires no labels, separate teacher, or reward signal, and adds no inference cost. On LLaVA-OneVision-2-8B, S$^3$T improves VSTAT accuracy by $+1.74$ as a single model, $+2.38$ with souping, and $+2.70$ with additional vision-encoder adaptation, while prior self-evolving methods leave state tracking largely unchanged. The capability learned from unlabeled synthetic clips transfers to real videos, improving performance by $+7.95$ on VSTAT-YouTube state-tracking questions and $+4.50$ on MVBench Action Count. --- ### TokenMatch: 3D Mesh Correspondence Transformer with Curvature-Guided Tokenisation Authors: Adeela Islam, Zorah Lähner, Vittorio Murino, Vladislav Golyanik Published: 2026-09-03 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.04202v1 While data-driven 3D shape correspondence estimation has recently seen substantial progress, robust matching under partial observations and strong non-isometric deformations remains challenging. Existing learning-based approaches often rely on hand-crafted descriptors or template-based representations, whereas recent generative models over functional maps suffer from high inference cost, limited interpretability, and poor generalisation to partial shapes. In response to these limitations, this paper introduces TokenMatch, a new transformer-based unified model for estimating 3D shape correspondences. Our feed-forward approach trained exclusively on BeCoS, a challenging non-isometric partial-to-partial shape-matching dataset, can generalise to matching full shapes without retraining or fine-tuning. TokenMatch uses self- and cross-attention mechanisms to efficiently learn patch-level and point-level relations as well as dense correspondences between shape pairs. Our core insight is that meshes can be adaptively tokenised into patches using shape curvature guidance, enabling effective learning of shape-specific geometric descriptors for correspondence estimation. We evaluate TokenMatch on standard benchmarks for partial and full shape matching, including CP2P, PSMAL, BeCoS, FAUST, SCAPE, and SHREC'19. Our method achieves consistently high performance, in most cases outperforming existing methods for partial and full shape matching in the mean geodesic error and intersection-over-union metrics, while also running faster at sub-second inference speeds. --- ### Scal3R: Learning Efficient Multi-Relative Pose Query for Scalable Online 3D Reconstruction Authors: Chin-Yang Lin, Yang-Che Sun, Cheng Sun, Fu-En Yang, Min-Hung Chen, Yen-Yu Lin, Wei-Chen Chiu, Yu-Lun Liu Published: 2026-09-03 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.04201v1 Online 3D reconstruction models perform poorly on long videos. This happens because regressing poses relative to a fixed first-frame anchor forces extrapolation far beyond the training distribution. Small drifts accumulate and amplify into significant geometric collapse. However, we observe that per-frame depth remains stable throughout this failure. The backbone's local geometry remains intact; only the global pose head breaks down. Motivated by this decoupling, we introduce Scal3R. This approach reformulates online reconstruction as multi-reference relative pose querying. We use lightweight learnable tokens, which make up about ~1% of the parameters, and inject them into a completely frozen backbone via asymmetric attention. This setup queries poses relative to multiple past keyframes. An online pose-graph optimization system with loop closure suppresses long-range drift. Scal3R reaches convergence in 8 hours on a single GPU. It reduces the average ATE by over 60% on KITTI compared to the online baseline. It also achieves state-of-the-art performance across Virtual KITTI, Sintel, TUM-Dynamic, ScanNet, and 7-Scenes. Project page: https://linjohnss.github.io/scal3r/ --- ### Principia: Relational Physics Tests for Video Models Authors: Varun Varma Thozhiyoor, Shivam Tripathi, Venkatesh Babu Radhakrishnan, Anand Bhattad Published: 2026-09-03 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.04200v1 Evaluating physical reasoning in video models is difficult because absolute motion measurements depend on frame rate, object scale, and camera calibration, all of which are often ambiguous or unavailable in generated video. We propose a different approach. When two objects in the same scene obey the same physical law, their motions must satisfy predictable relationships, and these relationships hold independent of calibration. We introduce Principia, a benchmark that evaluates Newtonian physics through relational consistency between paired objects. Principia spans eight phenomena - gravity, restitution, friction, rotational inertia, projectile motion, momentum, pendulum, and mass-spring oscillation - across translational, rotational, collisional, and oscillatory dynamics, using real-world scenes recorded under controlled protocols. We also introduce a calibration-independent consistency score that quantifies physical violation directly in image space. Across thousands of generations from six state-of-the-art video generators, no model exceeds 0.42 on Principia despite all scoring around 0.8 on VBench. Vision-language models are evaluated on their ability to detect relational physics violations, with the best model achieving only 67% accuracy and most performing near chance level. --- ### Compile by Training: Turning Natural-Language Specifications into Local Neural Functions Authors: Yuntian Deng, Pengyu Nie, Stuart Shieber Published: 2026-09-03 Categories: cs.CL, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2609.04199v1 Many recurring text functions are easy to describe but difficult to implement with rules, while calling a large remote model for every input introduces repeated cost, latency, and dependency on a provider. We present compile by training, which turns a natural-language specification into a reusable neural function. At compile time, teacher models generate task-specific examples that are used to train a small adapter for a compact interpreter. The resulting function runs without the teachers and can be stored, versioned, and composed like ordinary software. On FuzzyBench-Hard, a subset on which the Program-as-Weights fast compiler produced no exact matches, compile by training reaches 83.6% semantic accuracy. This higher accuracy comes with a higher compile-time cost: roughly a minute rather than seconds for the fast compiler. We deploy the compiler in a public interactive service and demonstrate compiled functions in a multi-site website helper, a language-controlled 3D avatar, and a bidirectional English-Claudish translator. --- ### Clean Engineering, Unstable Measurement: A Preregistered Reliability Failure of Black-Box LLM Observers on Shared Endpoints Authors: Haoyaun Zhu, Jie Zhang Published: 2026-09-03 Categories: cs.AI, cs.LG arXiv: https://arxiv.org/abs/2609.04198v1 Language-model judges now gate training data, score generations, and drive leaderboards. The judge is then a measurement instrument, resting on one rarely stated assumption: the same request, sent to the same model name, reads the same tomorrow. We audited that assumption in two preregistered campaigns with every threshold fixed in advance; neither got past validating its instrument. Across 52,988 audited request attempts, same-window repeat rankings agreed at Spearman 0.400 against a required 0.90, and byte-identical next-day replays agreed at 0.78 against a required 0.99, each time with the execution record at ceiling. Three mechanisms explain the gap: a label-to-meaning mapping that biased readouts as strongly as the signal; candidate gaps seven orders of magnitude below the instrument's own noise floor; and byte-identical inputs returning different rankings, a noise that exact-permutation readouts compound. Neither metric substitution nor sampling repaired it on the tested grid. Preregistered follow-ups bound the problem: waiting did not help on the days sampled (0.805 versus 0.800, replicated over five further days); switching providers did not help (four providers share the floor, medians 0.74 to 0.88, predicted by none of the metadata fields they expose); self-hosting on batch-invariant kernels helped only while the server was quiet; and on constructed errors with known gaps, the readout's separation tracks error type, not size. We distill the evidence into a three-level snapshot-identity ladder, eight design rules, and a reporting checklist; a pilot at roughly 2% of the study's call volume would have exposed both unreachable gates in advance. All results concern externally measured behaviour on shared serving infrastructure. On a shared endpoint, a model name is not a frozen instrument; a preregistered evaluation must measure its instrument before freezing any gate on it. --- ### ESPO: Error-Structured Prompt Optimization via Diagnose, Diversify, and Stabilize Authors: Lihao Liu, Peng Tang, Kunwar Yashraj Singh, Shabnam Ghadar Published: 2026-09-03 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2609.04197v1 Evolutionary prompt optimizers such as GEPA suffer from prompt bloat: each iteration appends rules and caveats, producing prompts up to 3$\times$ longer yet no more accurate. We trace this to three deficiencies - incomplete error observation, limited search diversity, and unreliable selection - and propose ESPO (Error-Structured Prompt Optimization), which decomposes prompt optimization into three phases: Diagnose clusters all training errors into structural patterns in one round; Propose generates candidates via four complementary strategies with independent biases; Select applies bootstrap stability selection. On seven public NLP benchmarks - Tweet, MMLU, GSM8K, HotpotQA, ScoNe, HoVer, and PUPA - ESPO improves average accuracy by $+$3.76 pp over the state-of-the-art (74.67% vs 70.91% for GEPA), matching or exceeding GEPA on every dataset while producing prompts 47% shorter (1,004 vs 1,878 chars) and faster at inference. Cross-model experiments across four additional student models (Gemma 3 12B, Mistral 14B, Qwen3 32B, Claude Haiku 4.5) show ESPO yields the best average accuracy on every model tested, with the largest gap on Qwen3 GSM8K (15.00% $\to$ 91.40%). A generalization bound (Appendix) grounds each phase in a corresponding term of the test-time gap, and the ablation confirms a key prediction: adding diversity without bootstrap selection actually hurts performance ($-$1.20%). --- ### Puffin-World: Scaling a Unified Multimodal Model with Native 3D World States Authors: Kang Liao, Yihang Luo, Xiao-Ming Wu, Linyi Jin, Size Wu, Chunyu Lin, Yao Zhao, Fei Wang, Wei Li, Chen Change Loy Published: 2026-09-03 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.04196v1 We propose Puffin-World, a unified multimodal architecture that integrates physical understanding, spatial simulation, and 3D world generation and reconstruction without relying on external offline modules. To reliably construct and interact with 3D worlds, our framework jointly models three native world states: physics (gravity field and latitude), geometry (depth), and appearance (image), together with a unified Omni-Camera representation that supports diverse tasks and flexible motions. Beyond modeling these states, we introduce a strategy for propagating physical dynamics across future frames. By grounding absolute camera properties in the real world, Puffin-World enables physically consistent and visually stable world generation. We further couple appearance and geometry within a single generative process, jointly synthesizing each future view and reconstructing its underlying geometry. This unified paradigm enables interleaved closed-loop applications requiring synergy across multiple tasks, including mimic and self-calibrated world exploration. To scale Puffin-World to complex scenarios, we construct Puffin-16M, comprising 15 million vision-language-camera triplets and 1 million trajectories featuring various and challenging motions. To foster further research in this area, we released the code, models, and datasets. --- ### Axonal delay dispersion decides whether a neuron detects an event or a sequence, and predicts cortical column diameter Authors: Cheng Bi, Jipeng Sun Published: 2026-09-03 Categories: q-bio.NC, cs.NE arXiv: https://arxiv.org/abs/2609.04195v2 Cortical neurons fire sparsely -- often fewer than one spike per sensory window -- making rate coding insufficient and temporal coding a necessity. That conduction delays convert firing order into synchrony is long established. What governs which class of temporal feature a neuron detects -- one volley of coincident input, or two in a particular order -- has not been examined. We propose a delay-signature framework in which the axonal conduction delays converging on a dendritic branch constitute a physical key: only input sequences whose spike-time differences the delays compensate arrive synchronously, and coincidence detection, via calcium plateau thresholds, converts that synchrony into an all-or-none output. In simulations of an integrator-neuron model we report three results. First, a single physical scalar -- the dispersion of the delay set -- moves a population from event detection to order-selective sequence detection. The transition is emergent under random delays and connectivity: at narrow dispersion sequence detectors do not exist, and the dispersion at which they overtake event detectors tracks the inter-event interval with a slope statistically indistinguishable from one. This maps a computational distinction onto the anatomical one between myelinated and unmyelinated projections, making myelination a switch on what a neuron computes, not only a regulator of speed. Second, the same dispersion sets the code's limits: it bounds the longest codable interval and fixes an absolute timing tolerance of about a millisecond, with slowing better tolerated than speeding. Third, that millisecond window and horizontal conduction velocity together predict cortical column diameter, and the two areas with direct measurements fall where the relation puts them. One anatomically measurable parameter thus sets what a neuron detects and the limits of what it can represent. --- ### Axonal delay dispersion decides whether a neuron detects an event or a sequence, and predicts cortical column diameter Authors: Cheng Bi, Jipeng Sun Published: 2026-09-03 Categories: q-bio.NC, cs.NE arXiv: https://arxiv.org/abs/2609.04195v1 Cortical neurons fire sparsely -- often fewer than one spike per sensory window -- making rate coding insufficient and temporal coding a necessity. That conduction delays convert firing order into synchrony is long established. What governs which class of temporal feature a neuron detects -- one volley of coincident input, or two in a particular order -- has not been examined. We propose a delay-signature framework in which the axonal conduction delays converging on a dendritic branch constitute a physical key: only input sequences whose spike-time differences the delays compensate arrive synchronously, and coincidence detection, via calcium plateau thresholds, converts that synchrony into an all-or-none output. In simulations of an integrator-neuron model we report three results. First, a single physical scalar -- the dispersion of the delay set -- moves a population from event detection to order-selective sequence detection. The transition is emergent under random delays and connectivity: at narrow dispersion sequence detectors do not exist, and the dispersion at which they overtake event detectors tracks the inter-event interval with a slope statistically indistinguishable from one. This maps a computational distinction onto the anatomical one between myelinated and unmyelinated projections, making myelination a switch on what a neuron computes, not only a regulator of speed. Second, the same dispersion sets the code's limits: it bounds the longest codable interval and fixes an absolute timing tolerance of about a millisecond, with slowing better tolerated than speeding. Third, that millisecond window and horizontal conduction velocity together predict cortical column diameter, and the two areas with direct measurements fall where the relation puts them. One anatomically measurable parameter thus sets what a neuron detects and the limits of what it can represent. --- ### Legibility is Not Interpretability: Comparing Judged and Actual Importance in Chain-Of-Thought Reasoning Authors: Kevin Du, Alexander Hoyle, Laura Ruis, Acyr Locatelli Published: 2026-09-03 Categories: cs.CL, cs.LG arXiv: https://arxiv.org/abs/2609.04194v1 Reasoning traces from chain-of-thought models appear to offer a legible window into how a model arrives at its answer. A growing body of work treats them as such, using LLM judges to diagnose errors, evaluate faithfulness, and provide step-level supervision via process reward models and generative critics. These practices rely on the text of a reasoning step carrying information about its functional role. But does the text actually encode information about which reasoning steps matter? We operationalize the importance of a reasoning step as its advantage: the change in expected reward, e.g., producing the correct final answer, from including that step, estimated via Monte Carlo rollouts. Basing ground truth on these estimates, we evaluate whether LLM judges can identify high-advantage steps and find that sufficiently capable LLMs can outperform a prevalence baseline but fall well short of a noise ceiling. Fine-tuning a model as a step-level critic yields strong improvement for incorrect responses but remains distant from ceiling for correct responses, suggesting that step importance is only partially recoverable from the text of the reasoning trace. Our findings contribute to a growing body of chain-of-thought faithfulness work that cautions against treating the legibility of reasoning traces as interpretability, especially with implications for process reward modeling. --- ### One Editor, Many Edits: A Unified Training-Free Framework for Diverse Video Editing Authors: Adheesh Sunil Juvekar, Onkar Kishor Susladkar, Kiet A. Nguyen, Muntasir Wahed, Nabeel Bashir, Xiaona Zhou, Tianjiao Yu, Vedant Shah, Ismini Lourentzou Published: 2026-09-03 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2609.04190v1 Video editing spans diverse editing paradigms, yet achieving high-quality instruction-guided and subject-guided editing within a single unified framework remains challenging. We introduce EditVid, a training-free framework combining sparse causal memory for local coherence, correspondence-based post-attention token injection for long-range identity preservation, and soft latent blending for edit locality. The same framework supports instruction-guided and reference-guided edits, including style transfer, attribute modification, object insertion, part-level editing, and subject replacement. On FiVE, EditVid achieves 78.16 FiVE-Acc, compared with 58.95 for the strongest evaluated training-free baseline, while obtaining competitive results on IVEBench. A user study further shows a 51.8\% overall preference for EditVid over 7 competing methods. --- ### Robust PAC Learning of Concurrent Stochastic Games Authors: Angel Y. He, David Parker Published: 2026-09-03 Categories: cs.LG, cs.GT, cs.LO, cs.MA arXiv: https://arxiv.org/abs/2609.04189v1 We introduce the first Probably Approximately Correct (PAC) learning framework for general-sum concurrent stochastic games (CSGs) with transition uncertainty, while addressing the challenge of Nash equilibrium (NE) existence. Our algorithm maintains data-driven $L^1$ confidence sets over transition kernels and solves a robust CSG to compute a social-welfare optimal $\varepsilon$-NE, using a robust MDP-based exploration mechanism to drive joint state-action coverage. Crucially, we introduce a Nash margin characterisation that enables principled reasoning about equilibrium existence: the framework either returns an $\varepsilon$-approximate NE whose social-welfare value is $\varepsilon$-close to optimal, or provides a sound certificate that no exact NE exists. Under a minimum reachability condition $p_{\mathrm{reach}}>0$ over relevant state-action pairs, the algorithm terminates after a polynomial number of trajectory samples, with sample complexity $\widetilde{O}\left( {R_{\max}^2 H^4 |S|^2 |A| / (p_{\mathrm{reach}} \varepsilon^2)} \right)$. Empirical results on benchmark CSGs demonstrate near-optimal performance, correct handling of equilibrium (non-)existence, and sample complexity consistent with theory. --- ### Seeing Before Synthesizing: VLM-Guided Transition Event Discovery for Weakly-Supervised Dense Video Captioning Authors: Ye-Chan Kim, Seunghee Choi, SeungJu Cha, Si-Woo Kim, Hwiseon Kim, Hyungee Kim, Dong-Jin Kim Published: 2026-09-03 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2609.04183v1 Weakly-Supervised Dense Video Captioning aims to localize and describe multiple events in untrimmed videos given only an ordered set of event-level captions per video. Recent work synthesizes auxiliary transition captions via LLM to provide additional vision-language alignment, but these captions lack visual grounding and are rigidly assigned to every inter-event gap at a fixed location and duration. To address these, we propose Seeing Before Synthesizing (SBS), a framework that adaptively provides visually grounded linguistic guidance only where warranted. Leveraging a VLM, we generate frame-level narratives for the inter-event gaps and detect transitions from the semantic variation across them. For identified transitions, we then refine inter-event temporal masks by blending the temporal midpoint with the semantic change point and selecting the width that maximizes vision-language alignment. Experiments on ActivityNet Captions and YouCook2 demonstrate state-of-the-art performance in both captioning and localization. --- ### Knowledge Acquisition During Pre-training? Large Language Models Learn Better With Auxiliary Views Authors: Joseph Lee, Yidi Huang, Dokyoon Kim, Shu Yang, Li Shen Published: 2026-09-03 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2609.04180v1 Gaps remain in our understanding of how large language models (LLMs) acquire knowledge during pre-training. We posit that auxiliary views, reformulations of knowledge, are causally helpful for learning. We design controlled experiments to isolate this. First, we confirm that repetition is necessary for acquisition and clarify that paraphrasing helps only at smaller batch sizes. Second, holding the token budget fixed, allocating tokens from document repetition to auxiliary views improves learning, counterintuitively, even for factual recall. Third, the effectiveness of auxiliary views is not contingent on the strength of the teacher model that generates them. Fourth, we identify forms of knowledge, contextual and foundational, that aid learning in the presence of prior knowledge gaps. Finally, we examine how these effects manifest mechanistically via layer-wise biases and compression. Together, our findings suggest that auxiliary representations of knowledge, which arise naturally in large pre-training corpora, are a key factor in the success of pre-training and offer a plausible explanation for why data diversity matters. --- ### A Computationally Feasible Framework for Causal Probabilistic Explanation Authors: Rafal Urbaniak, Sam Witty, Daniel Waxman, Andy Zane, Poorva Garg, Emily Bunnapradist, Sankaran Vaidyanathan, Jack Feser, Drew Lehe, Eli Bingham Published: 2026-09-03 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.04177v1 Explaining why a specific outcome occurred, and which inputs deserve the blame or credit, is central to philosophical, scientific, and policy analysis. Existing tools split into two camps. The theory of actual causality (AC) gives principled verdicts, but only for toy-sized models, because computing them requires enumerating counterfactual scenarios. Scalable attribution methods like SHAP (or even causal SHAP) at least partially ignore the causal structure that generated the data, and can give answers that conflict with a careful causal analysis. We close this gap with Probabilistic Causal Impact (PCI). PCI builds on actual causality and on Pearl's notions of probability of necessity and sufficiency, but recasts the question of explainability as an estimation problem on a probabilistic causal model that is easily approximated via Monte Carlo. By specifying a distribution over "candidate explanations," a distribution over counterfactual values, and a scoring function, PCI provides tractable, causally grounded, graded explanations, generalizing AC and Pearl's probability of causation as degenerate cases. We evaluate PCI in synthetic and real-world examples, spanning consistency checks with AC, scaling experiments, complex continuous-valued dynamical systems, and a real-world deployed causal machine learning model trained on millions of datapoints. --- ### Zero-Shot Novel Depth Synthesis Using 3D Foundation Models Scene Representations Authors: Denis M. Akola, David F. Fouhey Published: 2026-09-03 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.04174v1 3D Foundation Models (3DFMs) such as VGGT have recently pushed the boundaries of 3D vision by predicting rich unified representations with feed-foward transformers. The scene representations learned by these models enable strong performance on multiple 3D vision tasks. In this paper, we investigate using their internal representations to infer 3D in the scene from new views. Our hypothesis is that in order to solve the task of 3D reconstruction, these models need to learn a representation that includes a large amount of general knowledge about 3D scenes. After showing that it is possible to decode hidden surfaces from internal 3DFM representations, we propose a method, Z3D, that estimates pointmaps in unseen views by doing latent diffusion on 3DFM representation. We show that Z3D can predict realistic depth maps for new views across multiple datasets. --- ### Last Translation Benchmark Authors: Vilém Zouhar, Niyati Bafna, Mukund Choudhary, Maike Züfle, Sara Rajaee, Pinzhen Chen, Jannis Vamvas, Sara Papi, Ona de Gibert, Bhavitvya Malik, Eliya Habba, Orfeas Menis Mastromichalakis et al. Published: 2026-09-03 Categories: cs.CL arXiv: https://arxiv.org/abs/2609.04173v2 For scientific progress, we need benchmarks that test the limits of state-of-the-art models, and evaluation methods that inform us about failure cases. As models get stronger, standard benchmarks for machine translation are approaching saturation. Further, automatic translation metrics are unreliable, opaque, and vulnerable to reward-hacking. Even gold human evaluation is not problem-free, because it often lacks reproducibility, objectivity, and scalability. Overall, this prevents us from tracking progress in the field and identifying pathways for improvement. We introduce the Last Translation Benchmark, a collection of human-authored and peer-reviewed examples (texts, images, audio, videos) that break leading machine translation models. We also present a new evaluation approach: each example comes with handcrafted verification rules describing concrete failure cases on that example, therefore allowing reliable and actionable future evaluation. The Last Translation Benchmark is a live dataset that accepts ongoing contributions. The latest version is LTBv1, containing accepted contributions prior to September 1st 2026, with future releases planned as new data is continuously collected. --- ### Last Translation Benchmark Authors: Vilém Zouhar, Niyati Bafna, Mukund Choudhary, Maike Züfle, Sara Rajaee, Pinzhen Chen, Jannis Vamvas, Sara Papi, Ona de Gibert, Bhavitvya Malik, Eliya Habba, Orfeas Menis Mastromichalakis et al. Published: 2026-09-03 Categories: cs.CL arXiv: https://arxiv.org/abs/2609.04173v1 For scientific progress, we need benchmarks that test the limits of state-of-the-art models, and evaluation methods that inform us about failure cases. As models get stronger, standard benchmarks for machine translation are approaching saturation. Further, automatic translation metrics are unreliable, vulnerable to reward-hacking, and provide unactionable assessments. Even gold human evaluation is not problem-free, because it often lacks reproducibility, objectivity, and scalability. Overall, this prevents us from tracking objective progress in the field and identifying pathways for improvement. We introduce the Last Translation Benchmark, a collection of human-authored and peer-reviewed examples (texts, images, audio, videos) that break leading machine translation models. We also present a new evaluation approach: each example comes with handcrafted verification rules describing concrete failure cases on that example, therefore allowing reliable and actionable future evaluation. The Last Translation Benchmark is a live dataset that accepts ongoing contributions. The latest version is LTBv1, containing accepted contributions prior to September 1st 2026, with future releases planned as new data is continuously collected. --- ### Rethinking On-Policy Distillation of Large Language Models II: One Training Example Authors: Zixuan Fu, Bingxiang He, Yuxin Zuo, Haohuan Huang, Jinqian Zhang, Ruhang Xiao, Cheng Qian, Qinyu Luo, Huan-ang Gao, Yudong Wang, Zhiyuan Liu, Ning Ding et al. Published: 2026-09-03 Categories: cs.AI, cs.CL arXiv: https://arxiv.org/abs/2609.04172v1 On-policy distillation (OPD) combines student-generated rollouts with dense token-level supervision from a teacher. Existing work has mainly studied its algorithmic behavior, leaving the role of training data unclear. We examine this role at the data-minimal limit by training on a single query. One-shot OPD keeps improving for hundreds of steps and recovers most of full-data OPD's gain across task domains and model families. We explain this result through the states visited during training and the rate at which the student aligns with the teacher. We measure \emph{state coverage}, the fraction of the states full-data OPD visits that a query set's rollouts reach. A single query already reaches \(71.5\%\), most of it within the first 100 steps. Adding semantically distinct queries raises coverage and validation accuracy together, until 16 queries reach \(98.9\%\) and match full-data training. Yet alignment slows at a similar pace whether OPD trains on one query or the whole dataset, and even a fixed set of states takes hundreds of steps to absorb. OPD is therefore data-overfed but algorithm-starved. Its rollouts quickly expose broad supervision, while the student absorbs that supervision increasingly slowly. The state-coverage result extends to multi-teacher OPD, where 16 semantically diverse queries per domain match full-data MOPD. As a further stress test, content-light templates and off-domain WildChat queries also approach the real-query baseline. Task content and induced state coverage can therefore come apart. We hope these findings direct future work toward the step efficiency of OPD, and prompt a re-examination of the data and the mechanisms behind its recent successes in frontier post-training. --- ### A Case Study on Emergent Cheating and Whistleblowing in Autonomous Research Swarms Authors: Davide Paglieri, Logan Cross, Tim Genewein, Joel Z. Leibo, Nenad Tomasev, Alexander Sasha Vezhnevets Published: 2026-09-03 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.04170v1 Multi-agent AI science ecosystems rely on agents possessing tools that allow them to communicate, coordinate, and build on each other's work. Yet this shared infrastructure can also introduce vulnerabilities by creating a substrate for the contagious spread of unintended and undesirable behaviors. We report a case study on a research collective of 100 autonomous LLM agents tasked with proving formal mathematical conjectures. Within the swarm, cheating spontaneously emerged and was later challenged by whistleblowers - both without any external intervention. When a single agent discovered an exploit in the evaluation system, it propagated across the collective via a shared knowledge library and later through peer-to-peer messages. Despite early reluctance, a cohort of agents adopted the exploit in response to competitive pressure. A separate group of agents produced an emergent counter-response: auditing fraudulent proofs, alerting peers across broadcast and private channels, staging boycotts, lodging formal complaints, and proposing validation patches. In recent incidents, agent swarms coordinated covertly through improvised side-channels (Dalton and Wallace, 2026; Greenblatt et al., 2026). Our setting differs: the same transparent channels that carried the exploit also gave non-cheating agents the visibility they needed to detect fraud, organize resistance, and enforce norms. We cast the problem of managing the agents' shared infrastructure as the knowledge commons governance problem (Ostrom, 1990). To protect the commons from exploits, we propose to adopt institutional mechanisms, such as graduated sanctioning and collective-choice rules, to support decentralized self-governance in autonomous swarms. --- ### Para-Pipe: Exploiting Hierarchical Operator Parallelism of ML Computational Graphs on SoCs Authors: Yujie Zhang, Huiying Lan, Ehsan Aghapour, Zhiyuan Ning, Peng Zan, Weidong Shao, Anuj Pathania, Tulika Mitra Published: 2026-09-03 Categories: cs.DC, cs.LG, cs.PF arXiv: https://arxiv.org/abs/2609.04168v1 As edge-based deep learning applications become more complex, optimizing performance on heterogeneous System-on-Chips (SoCs) presents unique challenges. Traditional pipelining techniques distributing the computation across different on-chip processing units, while effective for throughput, do not address the latency demands posed by modern neural networks with complex interdependencies and extensive operator parallelism. There is a potential in leveraging operator parallelism to enable concurrent execution across multiple processing units, thereby reducing inference latency. However, prioritizing pipelining or parallel execution often necessitates a compromise, where optimizing one performance metric adversely impacts the other. This paper introduces Para-Pipe, a hierarchical mapping framework that integrates intra- and inter-stage operator parallelism within a pipelined architecture. Para-Pipe navigates the trade-off between throughput and latency by selectively fine-tuning parallelism levels within and across pipeline stages. This strategy can significantly reduce inter-processor communication overhead, significantly improving energy efficiency. Our evaluation demonstrates that Para-Pipe generates multiple Pareto-optimal configurations, achieving a balance between throughput and latency on an Amlogic SoC equipped with ARM big.LITTLE CPUs and GPU, as well as the Black Sesame Technology SoC featuring a deep learning accelerator and two DSPs. More importantly, throughput-optimized configurations under Para-Pipe on Amlogic SoC show an average energy efficiency improvement of 11.0% over purely pipelined strategies and 23.3% relative to non-pipelined parallel execution. --- ### SWE-Gate: Passing Functional Tests Is Not Enough for Software Engineering Agents Authors: Xin He, Yanlin Wang, Mingwei Liu, Jiachi Chen, Hongyu Zhang, Guanbin Li Published: 2026-09-03 Categories: cs.SE, cs.AI arXiv: https://arxiv.org/abs/2609.04167v1 Repository-level software engineering benchmarks have significantly advanced the evaluation of coding agents, but existing benchmarks primarily measure whether generated patches pass functional tests and overlook review-derived acceptance constraints (review constraints) that often influence whether a patch is acceptable in real-world software development. We introduce SWE-Gate, a repository-level benchmark for software engineering agents that explicitly evaluates review constraint compliance alongside functional correctness. SWE-Gate derives review constraints from real pull request review comments and synthesizes repository-level repair instances around these constraints. Each instance provides separate functional and constraint tests, together with non-compliant and gold patches, enabling explicit separation between issue resolution capability and review constraint compliance. We construct SWE-Gate with 303 repository-level repair instances spanning 75 open-source Python repositories across diverse software domains. Experiments with four LLM backends spanning different capability levels under a common coding-agent scaffold reveal a substantial gap between functional success and success under the complete repair specification: among 644 repairs that pass the functional tests, 221 fail to satisfy the provided review constraints. These findings show that functional-only evaluation overestimates agents' ability to satisfy the full requirements of repository-level repair tasks. The replication package including code, data, and experimental results is available at https://github.com/DeepSoftwareAnalytics/SWE-Gate. --- ### From Deceptive Outputs to Deceptive Mechanisms: A Causal Framework for Language-Model Deception Research Authors: Yakov Pyotr Shkolnikov Published: 2026-09-03 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.04166v1 Research and news coverage of language-model deception increasingly attributes human-like mental-state concepts to language models. Such claims can blur the distinction between behavior that looks deceptive and a mechanism that is actually deceptive. We introduce a causal taxonomy separating prior commitment from retrospective report, model preference from realized output, false preference from sensitivity to the utility of misleading a recipient, and deceptive behavior from the provenance of the objective or strategy producing it. We test these distinctions in two open-weight model families. Across controlled guessing-game and stock-trading experiments, we find that deceptive-looking behavior can arise without the corresponding proposed mechanism, while other interventions provide direct evidence that recipient information state can causally affect deceptive preference. These results show that deceptive behavior can provide evidence for a deceptive mechanism. But even evidence for such a mechanism does not establish model agency in the deception. --- ### Parameterised graph theory for tensor networks: entanglement rerouting, structural simplification, and agnostic tomography Authors: Matthias C. Caro, Natalie McHugh, Sergii Strelchuk Published: 2026-09-03 Categories: quant-ph, cs.DS, cs.LG arXiv: https://arxiv.org/abs/2609.04165v1 Parameterised graph theory studies how the complexity of graph-theoretic problems depends on structural parameters of the input graph. This perspective has proved useful in analysing tensor-network simulation (Markov and Shi, 2008). Its implications for tensor-network representations and tomography are less well understood. In particular, which graph parameters determine whether a tensor-network state (TNS) admits a tractable matrix product state (MPS) or tree tensor network (TTN) representation, and which control the complexity of learning the state? We address these questions using parameterised graph theory. First, we show that cutwidth and tree-cutwidth bound the bond dimension overhead required to represent a TNS as an MPS or TTN. In the TTN case, tree-cutwidth also bounds the local dimension of the grouped subsystems. The proofs are based on entanglement rerouting, a tensor-network analogue of rerouting information in a classical network. Second, we derive graph-dependent upper bounds on the sample and computational complexity of realisable TNS tomography, with exponents that depend on cutwidth, tree-cutwidth, and a new graph parameter, learning complexity, which we bound in terms of degree and treewidth. We obtain these results by extending the disentangling MPS learner of (Cramer et al., 2010), as analysed further in (Bakshi et al., 2025; Lin et al., 2025), to TTNs and to tensor networks on arbitrary known graphs. Finally, we extend the framework beyond the realisable setting. For an arbitrary input state, our agnostic learner outputs a pure state whose fidelity is within additive error $ε$ of the optimum over tensor-network states on the given graph with a given bond dimension, with explicit graph-dependent bounds on sample and computational complexity. --- ### SENTINEL-RL: Offloading Topological Reasoning from LLM Agents in the Security Operations Center Authors: Uday Vallabhaneni, Cassie L. Cagwin, David J. Wild Published: 2026-09-03 Categories: cs.CR, cs.AI arXiv: https://arxiv.org/abs/2609.04159v1 Large language model (LLM) agents are increasingly proposed as autonomous SOC analysts, but two limitations make them unreliable at enterprise scale: a finite context window cannot hold a multi-thousand-host authentication graph, and free-form generation offers no guarantee that a recommended containment action is consistent with the topology it operates on. We present Sentinel-RL, an agentic-SOC architecture that decouples topological reasoning from semantic reasoning: a heterogeneous graph attention encoder summarizes the live authentication subgraph into a fixed-dimensional state, a Proximal Policy Optimization (PPO) policy maps this state to a constrained set of investigative actions, and an LLM agent loop is restricted to consuming the policy's recommendations and producing analyst-readable narratives gated by a critic. We instantiate the system on the LANL Comprehensive, Multi-Source Cyber-Security Events dataset and the Indiana University Quartz HPC cluster, reporting four results: (i) a two-phase CREATE ingestion pattern loads a 24M-edge authentication subgraph into Neo4j in 14.2 minutes on a single 32-core node, roughly 24x faster than the canonical MERGE-based pipeline; (ii) a sliding-window alert engine reliably trips a 25-event/10-second threshold in <=2.5 s across 50 trials; (iii) PPO training over 200 iterations converges to a mean episodic return of 8.74+/-0.31, with held-out precision of 0.91 and recall of 0.87 on labeled red-team events; and (iv) the integrated containment loop completes a full detect-investigate-recommend-human-approve cycle in a median of 6.3 s. We contribute a reusable engineering pattern (the hot-node deadlock workaround), a portable HPC deployment pattern (anchor-node co-location), and an enterprise-readiness analysis covering false-positive economics, reversibility guarantees, audit compliance, and the human-approval boundary. --- ### Persistent Identity Preservation in Generative Image Models: A Benchmark and Evaluation System Authors: Mengwei Ren, Xuaner Zhang, Zhihao Xia Published: 2026-09-03 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.04151v2 Generative image models can now produce high-quality images, follow complex instructions, and support precise edits, but they still struggle to preserve who or what is being depicted. When generating or editing images of a specific subject, identity may drift as the pose, expression, appearance, viewpoint, or surrounding scene changes. Existing subject-driven methods make fundamentally different choices about where identity is represented: through the input context (GPT-Image-2, NB2), as trainable subject-specific model parameters (LoRA), or as a persistent identity layer (PHOTA IDENTITY) reusable across generations and edits. We systematically benchmark these paradigms across subject-driven generation, editing, restoration, and multi-subject settings, with tasks designed to increasingly stress identity preservation. Our results show that identity preservation remains a distinct limitation of current generative foundation models: strong image quality and instruction following do not necessarily imply strong identity fidelity, and identity degradation becomes more pronounced under iterative edits, small subject scales, severe image degradation, and multi-subject composition. Persistent identity substantially reduces this degradation across generation, editing, and restoration, consistently improving identity preservation when applied to different foundation models while maintaining comparable instruction adherence and perceptual image quality. These results suggest that identity does not simply emerge from increasingly capable generative models, but can instead be represented as persistent subject knowledge that is composed independently with the underlying generative model. --- ### Persistent Identity Preservation in Generative Image Models: A Benchmark and Evaluation System Authors: Mengwei Ren, Xuaner Zhang, Zhihao Xia Published: 2026-09-03 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.04151v1 Generative image models can now produce high-quality images, follow complex instructions, and support precise edits, but they still struggle to preserve who or what is being depicted. When generating or editing images of a specific subject, identity may drift as the pose, expression, appearance, viewpoint, or surrounding scene changes. Existing subject-driven methods make fundamentally different choices about where identity is represented: through the input context (GPT-Image-2, NB2), as trainable subject-specific model parameters (LoRA), or as a persistent identity layer (PHOTA IDENTITY) reusable across generations and edits. We systematically benchmark these paradigms across subject-driven generation, editing, restoration, and multi-subject settings, with tasks designed to increasingly stress identity preservation. Our results show that identity preservation remains a distinct limitation of current generative foundation models: strong image quality and instruction following do not necessarily imply strong identity fidelity, and identity degradation becomes more pronounced under iterative edits, small subject scales, severe image degradation, and multi-subject composition. Persistent identity substantially reduces this degradation across generation, editing, and restoration, consistently improving identity preservation when applied to different foundation models while maintaining comparable instruction adherence and perceptual image quality. These results suggest that identity does not simply emerge from increasingly capable generative models, but can instead be represented as persistent subject knowledge that is composed independently with the underlying generative model. --- ### Terminal-Universe: Turning Agent Trajectories into Scalable Terminal Environments Authors: Jie Wu, Zhenru Zhang, Beichen Zhang, Xuwu Wang, Yuhui Su, Mouxiang Chen, Peng Wang, Zhihai Wang, Que Shen, Hao Zhou, An Yang, Fei Huang et al. Published: 2026-09-03 Categories: cs.AI, cs.CL arXiv: https://arxiv.org/abs/2609.04148v1 As terminal-based code agents become prevalent, agent trajectories have accumulated at scale, while realistic, executable environments remain scarce. However, environments are what agent post-training actually requires: each can be re-queried into many verifiable tasks and provides execution feedback, whereas a trajectory is a single frozen demonstration. Rather than generating environments from scratch, we observe that the tool-execution history in existing trajectories exposes the structure and contents of the environments in which they ran, making it possible to reconstruct those environments from the trajectories themselves. Thus, we introduce Terminal-Universe, a framework which turns each trajectory into a reusable environment and explores it for synthesizing new tasks and continued interactions. Specifically, Terminal-Universe replays the file operations recorded in a trajectory to restore each file before the agent modified it, yielding a partial workspace; a completion agent then supplies the missing files and dependencies. On this recovered workspace, we both reconstruct the original intent task and synthesize entirely new ones. Besides, we also scale the tasks along two complementary axes: breadth and depth. For breadth, we mine directional dependency relations between related environments and synthesize cross-workspace queries spanning multiple codebases, as developers routinely do in real-world development. For depth, we extend the initial single-turn query into a multi-round session that captures iterative user feedback and requirement refinement via a user agent. Applied to public terminal agent trajectories, Terminal-Universe produces 37.3k task-sufficient environments. Supervised fine-tuning of Qwen3.5-27B on this corpus improves single-round performance on Terminal-Bench 2.1 by 11.9 points and multi-round performance on EvoCode-Bench v2 MT@4 by 13.8 points. --- ### A Low-Cost, Open Platform for End-to-End Autonomous Driving on a Miniature Ackermann Vehicle Authors: Gustavo Claudio Karl Couto, Eric Aislan Antonelo, Gabriel George Zipperer Published: 2026-09-03 Categories: cs.LG, cs.AI, cs.RO arXiv: https://arxiv.org/abs/2609.04147v1 This paper presents a low-cost, open experimental platform for research in end-to-end autonomous driving with miniature Ackermann vehicles. The platform combines a physical vehicle, a printed urban track, data collection tools, trajectory registration, and a Webots digital twin, enabling controlled experiments that connect simulation-based autonomous-driving methods to real-world execution. As a first baseline, we implement command-conditioned behavior cloning, in which a neural policy receives an on-board camera image and a high-level navigation command and outputs steering and speed. The system is evaluated both on the physical vehicle and in simulation. In real closed-loop experiments, the learned policy follows lanes and executes commanded turns, reaching a mean cross-track error of 6.1 cm with respect to the reference route, close to the 4.7 cm observed in human demonstrations. In the digital twin, camera field of view has a strong effect on performance, reducing the mean cross-track error from 35.6 to 3.3 cm when widened from 58 to 120 degrees. Using the digital twin to generate synthetic driving data and a learned sim-to-real image translator to reduce the appearance gap, we further show that a higher-capacity policy trained on this synthetic data combined with real demonstrations is the only configuration that completes all four track routes in closed loop, whereas the compact baseline and the same network trained on real data alone complete fewer. These results establish the open platform as a practical testbed for sim-to-real studies and provide an initial command-conditioned imitation-learning baseline; we release it to support reproducible research. --- ### Efficient Test-Time Adaptation through Human-AI Interaction Authors: Zora Zhiruo Wang, Apurva Gandhi, Rulin Shao, Aspen Chen, Jonas Mueller, Zhiqi Liang, Jett Chen, Michael Ryan, Qianou Ma, Luxi He, Zhoujun Cheng, Andre He et al. Published: 2026-09-03 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.04141v1 AI agents are trained on population-scale data to encode broad capabilities spanning those of many practitioners. Yet the artifacts they produce rarely meet the personal bar professionals need to stake their reputation on. On realistic, open-ended tasks where success criteria are heterogeneous and insufficiently documented, individual expertise lives precisely in the elevation and departure from the average. In practice, iterative human-agent interaction surfaces criteria that users cannot fully specify up front, yet apply repeatedly across tasks. We argue this cross-session interaction data is a rich, underused signal for closing the gap to individual expertise. In this work, we propose test-time adaptation through human-agent interaction (TAHI), which integrates these signals into agent context and weights, and crystallizes each user's training and evaluation criteria via an evolving rubric module. We adapt agents to 30 individuals in two high-utility domains, writing and visual creation, on a total of 600 tasks. Our agents improve solo task success by 4.5-20.9% within only tens of tasks. Meanwhile, our evolving rubric module serves as a scalable annotation tool, creating evaluation rubrics that catch 16.0-22.3% more failures than those from LMs or humans alone. While agents are adapted towards individuals, we show these personalized agents also produce improvements in success of up to 8.8% that generalize across users. --- ### The Natural Language Interaction Protocol and Standard for AI Agents Authors: Luyi Xing, Rasit Onur Topaloglu, Ranjan Sinha, Abhay Ratnaparkhi, Samuel Ndichu, Christopher Nguyen, Anindita Das, Tom Sheffler, Mohamed Rahouti, Zichuan Li, Xiaojing Liao, Sanjay Aiyagari Published: 2026-09-03 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.04135v1 AI agents are increasingly being developed and deployed across organizations using heterogeneous agent-development frameworks, AI models, tool interfaces, protocols, and execution environments. To realize their potential social and business impact, these agents must be able to interoperate through a common communication protocol. The Natural Language Interaction Protocol (NLIP), developed by researchers and practitioners across companies and universities and standardized by Ecma International, addresses this need by defining a standards-based application-layer protocol for AI-agent interaction. NLIP provides a lightweight semantic message envelope that can be carried over existing transports such as HTTP/HTTPS, WebSocket, and AMQP, while allowing NLIP-aware agents and gateways to adapt between clients, agents, local context stores, ontologies, tools, enterprise services, and heterogeneous underlying protocols. This paper presents the motivation and design rationale of NLIP, its message model and transport bindings, security-by-design considerations, reference implementation, representative applications, adoption signals, and relationship to emerging agent protocols such as MCP and A2A. --- ### Prospective Coding Improves Learning in Deep Continuous-Time Recurrent Networks Authors: Shivang Rawat, Mirko Morello, Flaviano Morone, David J. Heeger Published: 2026-09-03 Categories: cs.LG, cs.NE, q-bio.NC arXiv: https://arxiv.org/abs/2609.04134v1 Temporal integration gives continuous-time recurrent networks memory, but in deep stacks it also delays bottom-up signals and attenuates top-down errors. We develop Recursive Quadrature Filters (RQFs), biologically motivated complex-valued temporal filters that are a special case of diagonal state-space models (SSMs), and ask whether this failure mode can be addressed by making each layer's bottom-up input prospective. Starting from an energy model, we derive the RQF dynamics and show that each RQF is a band-pass filter whose learnable parameters control its tuning frequency and bandwidth. We then make each layer's bottom-up input prospective using a parameter-free two-tap update that leaves the recurrent transition and parallel scan unchanged. We extend this correction to general diagonal SSMs and show that it mitigates depth-dependent gradient attenuation when temporal gradients are truncated, i.e., spatial-only backpropagation. We evaluate the intervention in RQFs, S5, and ORGaNICs (a nonlinear gated RNN) trained using full backpropagation through time (BPTT) and spatial-only backpropagation. Under full BPTT, prospective variants match or outperform their non-prospective controls in every model and configuration. A non-residual width-32 six-layer RQF reaches 96.09% accuracy on raw-audio Speech Commands with 31.9k parameters; a width-64 six-layer RQF reaches 83.56% on the 16,384-step Path-X task. These results identify RQFs as a parameter-efficient recurrent substrate and prospective-input coding as an input-side correction for deep continuous-time recurrent networks. --- ### Beyond Retrieval: Progressive Latent Memory Evolution for Streaming Video Understanding Authors: Hongyu Qu, Guangming Yao, Ling Xing, Xiaobin Hu, Rongxing Ding, Guibin Zhang, Fan Zhang, Yi Yuan, Xiangbo Shu, Shuicheng Yan Published: 2026-09-03 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.04131v1 Streaming video understanding requires multimodal large language models (MLLMs) to process continuous visual inputs and respond to user queries under strict causality and bounded memory. Existing approaches typically compress historical observations into an external memory bank and retrieve query-relevant evidence as additional visual context. Though effective, this store-and-retrieve paradigm keeps historical evidence as external visual context, preventing it from being internalized into a compact, evolving latent memory that can continuously guide streaming reasoning. To bridge this gap, we introduce LatentStream, a progressive latent working memory framework that shifts streaming memory from store-and-retrieve to retrieve-and-internalize. Specifically, LatentStream comprises three coordinated components. First, Query-agnostic Hierarchical Streaming Memory organizes visual history into short-, mid-, and long-term levels under a fixed memory budget through Jenks-guided adaptive consolidation. Once a query arrives, Hierarchical Latent Memory Evolution equips groups of latent memory tokens with progressively expanding memory receptive fields, enabling them to iteratively retrieve historical evidence from their corresponding scopes and internalize it into a compact, fixed-length latent memory. Finally, Progressive Confidence-guided Latent Memory Optimization constructs a hierarchical progression reward from group-wise predictive entropy and jointly refines the latent memory tokens and retrieved evidence, encouraging increasingly confident streaming reasoning. Extensive experiments demonstrate that LatentStream achieves new state-of-the-art results on existing online and offline video benchmarks. --- ### Environment Evolution for Terminal Agents Authors: Zhiyuan Fan, Tinghao Yu, Yuanjun Cai, Jiang Zhou, Jiangtao Guan, Jincheng Liu, Yun Yang, Dingxin Hu, Zhuo Han, Xing Wu, Feng Zhang, Lilin Wang Published: 2026-09-03 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.04128v1 Scaling interactive and verifiable environments is critical for training terminal agents. As frontier models become more capable, environments synthesized from scratch become less challenging and thus provide limited learning signals. Recent co-evolution methods iteratively synthesize environments near the model's learnable frontier based on weaknesses exposed during rollouts. However, their dependence on on-policy rollouts limits generalization and the continuous provision of learning signals as the model becomes stronger. In this paper, we propose environment evolution, which incrementally increases environment difficulty off-policy and schedules the evolved environments generation by generation during training to provide continuous learning signals. We derive three evolution directions that influence environment difficulty from the multi-turn learning objective and then implement evolution along these directions through a loop-engineered multi-agent harness. Quantitative rollout experiments with Hy4 preview, Claude Opus 5, and GPT-5.6 Sol show that environment evolution consistently produces more difficult environments. We validate its effectiveness on Qwen3.6-27B and Qwen3.6-35B-A3B through simple long-horizon RL training, improving their performance by 14.4 and 18.0 percentage points on Terminal-Bench 2.1, respectively. --- ### Epistemic Warrant for LLM Recommendations: Characterizing the Basis for Reliance When Ground Truth Is Unavailable Authors: Shai Vardi, João Sedoc Published: 2026-09-03 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.04127v1 Large language models are increasingly used to support organizational decisions, yet users often lack a principled basis for assessing whether to rely on a specific recommendation. Existing approaches typically evaluate broad model properties, such as reliability, uncertainty, or robustness, or focus on user trust, rather than the underlying basis for relying on an individual recommendation. Adapting theoretical foundations from epistemology, we introduce epistemic warrant, a decision-level construct that characterizes the stability of a model's preference and the scope over which that preference holds. We operationalize this construct through a four-tier reliance certificate for pairwise recommendations, distinguishing among unstable, context-dependent, locally supported, and broadly supported recommendations. We validate the construct using contemporary methodologies: known-groups tests successfully recover expert-prespecified warrant orderings, and stronger warrants systematically align with independent consensus from crowd workers. Furthermore, we demonstrate that epistemic warrant provides information distinct from verbalized confidence and is not readily explained by decision difficulty. Ultimately, this framework offers a theoretically grounded, implementable approach for characterizing the warrant of individual LLM recommendations when objective ground truth is unavailable. --- ### BooM-VVT: Boosting Mask-Free Video Virtual Try-On with Image-Level Pseudo Data Authors: Wei Zhang, Xin Li, Peishu Shi, Jialin Gao, Xuekang Peng, Zhichao Lian, Yeying Jin Published: 2026-09-03 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.04120v1 Video virtual try-on (VVT) aims to generate realistic videos of a person wearing a target garment. Recent methods leverage a keyframe-driven video generation paradigm to improve in-the-wild performance, yet they still rely on masks to localize try-on regions, making them vulnerable to large motions and severe occlusions. Although mask-free image-based try-on methods have shown promising results by leveraging large-scale pseudo data, extending this paradigm to videos remains difficult, as constructing video-level pseudo data is prohibitively expensive. Furthermore, coarse keyframe sampling and the scarcity of multi-view try-on data limit existing keyframe-driven methods in maintaining garment consistency and handling diverse try-on tasks. To address these challenges, we propose BooM-VVT, a mask-free VVT framework built upon the keyframe-driven paradigm. To achieve mask-free VVT, we introduce a multi-stage training strategy that leverages image-level pseudo data for mask-free localization learning, substantially reducing the need for costly video-level pseudo data. To improve garment consistency, we propose Garment-Sensitive Keyframe Sampling, which selects keyframes based on garment-relevant body regions to better capture garment appearance. We further introduce Frame-Shared 3D-RoPE to establish spatiotemporal correspondences between keyframes and target video frames for accurate garment-detail transfer. Finally, we construct OmniView, a large-scale multi-view try-on dataset to support reliable try-on video generation under complex camera viewpoints and diverse try-on tasks. Extensive experiments demonstrate that BooM-VVT achieves superior temporal consistency and garment fidelity over existing methods. Project page: https://boomvvt.github.io/boomvvt. --- ### Constant regret in general games via higher-order optimism Authors: Omar Abbadi, Rida Laraki, Panayotis Mertikopoulos Published: 2026-09-03 Categories: cs.LG, cs.GT arXiv: https://arxiv.org/abs/2609.04113v1 We introduce an uncoupled learning algorithm which, when employed by all players of an arbitrary $N$-player normal form game with up to $K$ actions per player, guarantees $O(N^3\log^2 K)$ individual regret, uniformly over the horizon of play. The proposed algorithm - which we call higher-order optimism with discounting (HOOD) is a variant of optimistic follow-the-regularized-leader (OptFTRL) that combines a discounted $(N+1)$-th order predictor with entropic regularization over a suitable "lifting" of the game's strategy space. This combination of ingredients is purposefully designed to dampen large oscillations of the induced sequence of play in a controlled manner, removing in this way a key stumbling block of previous attempts to achieve constant regret in general games. Our approach bears several striking similarities to the concurrent - and completely independent - work of Liu, Farina, and Ozdaglar (arXiv:2608.31166), who very recently derived an $O(N^{21}\log^{4} K)$ regret bound through the use of higher-order optimism and an exponential moving average estimator. --- ### The Shape of Time: Video-Token Contrast for Temporal Understanding in VideoLMs Authors: Yumeng Shi, Quanyu Long, Yin Wu, Wenya Wang Published: 2026-09-03 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.04110v1 Seeing frames in order does not mean representing time. Modern VideoLMs receive ordered video streams, yet their main supervision acts on generated text rather than video-token representations where event dynamics should first emerge. This mismatch allows models to learn temporal answers from shortcuts such as objects, scenes, and language priors, without requiring internal video representations to capture event progression. To address this, we propose VT-Contrast, a representation-level temporal counterfactual objective for VideoLMs. Its design asks where temporal supervision should act and what temporal differences it should expose. VT-Contrast supervises selected late-layer last-frame video tokens, where temporal information is expected to be integrated before language generation, and contrasts order-preserving views with same-video reordered counterfactuals graded by Kendall tau distance. It requires no architectural changes, is compatible with diverse VideoLM training tasks, and improves overall performance across temporal understanding benchmarks. Our code is available at https://github.com/ANDgate99/VT-Contrast. --- ### Sequential Beats Joint: On the Interplay between On-Policy Distillation and RLVR Authors: Boyan Li, Bingsen Chen, Chenghao Yang, Ping Nie, Chen Zhao, Xi Ye Published: 2026-09-03 Categories: cs.CL, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2609.04108v2 Reinforcement learning with verifiable rewards (RLVR) and on-policy distillation (OPD) have emerged as two dominant methods for post-training reasoning LLMs. Prior work uses OPD's dense token-level supervision to complement the sparse RL reward, fusing the two signals within a single step: either as a \emph{weighted-additive combination} or a \emph{teacher-modulated rescaling} of the RL advantage. In this paper, we show that a simple two-stage scheme, OPD-then-RL, consistently outperforms pure OPD, pure RLVR, and all such joint baselines across logic and math reasoning benchmarks. Beyond the empirical results, we further provide a systematic understanding of this through pass@$k$ behavior, learning dynamics, and parameter updates, yielding a consistent explanation: OPD expands the student's coverage of teacher-supported solutions and RL sharpens within that support, while jointly optimizing the two signals causes them to interfere. To provide a practical recipe, we find that the OPD validation score is the key signal for when to switch to RL, and that OPD is a better cold start for RL than SFT. Together, our results establish OPD-then-RL as a simple yet strong way to combine the two methods, turning two entangled signals into complementary stages. --- ### Sequential Beats Joint: On the Interplay between On-Policy Distillation and RLVR Authors: Boyan Li, Bingsen Chen, Chenghao Yang, Ping Nie, Chen Zhao, Xi Ye Published: 2026-09-03 Categories: cs.CL, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2609.04108v1 Reinforcement learning with verifiable rewards (RLVR) and on-policy distillation (OPD) have emerged as two dominant methods for post-training reasoning LLMs. Prior work uses OPD's dense token-level supervision to complement the sparse RL reward, fusing the two signals within a single step: either as a \emph{weighted-additive combination} or a \emph{teacher-modulated rescaling} of the RL advantage. In this paper, we show that a simple two-stage scheme, OPD-then-RL, consistently outperforms pure OPD, pure RLVR, and all such joint baselines across logic and math reasoning benchmarks. Beyond the empirical results, we further provide a systematic understanding of this through pass@$k$ behavior, learning dynamics, and parameter updates, yielding a consistent explanation: OPD expands the student's coverage of teacher-supported solutions and RL sharpens within that support, while jointly optimizing the two signals causes them to interfere.To provide a practical recipe, we find that the OPD validation score is the key signal for when to switch to RL, and that OPD is a better cold start for RL than SFT. Together, our results establish OPD-then-RL as a simple yet strong way to combine the two methods, turning two entangled signals into complementary stages. --- ### Hardware-Aware FP4 FlashAttention-4 Authors: Robert Hu Published: 2026-09-03 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.04105v1 Blackwell's 4-bit floating-point (FP4) tensor cores do not automatically make attention faster because softmax conversion and on-chip dependencies dominate once its matrix products shrink. We address this with \emph{Direct-P} for noncausal inference and a causal path that passes the forward quantization directly into backward. Direct-P maps scores directly to FP4 probabilities and reaches up to 2.13$\times$ the bfloat16 (BF16) forward throughput on an NVIDIA GB200. The causal path reconstructs probabilities from saved quantized queries and keys and uses 8-bit floating-point (FP8) gradient operands, accelerating a complete single-GPU 8-billion-parameter update by up to 1.14$\times$. Matched distributed training retains FP8 probabilities and values; every tested MXFP4 probability/value training trajectory diverges. --- ### Why Gated DeltaNet Survives 4-Bit Quantization: NVFP4 W4A4 for the Recurrent Half of a Hybrid 27B LLM Authors: Sergii Kozyrev, Davyd Maiboroda Published: 2026-09-03 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.04098v1 Hybrid LLMs pair softmax attention with linear-attention layers such as Gated DeltaNet (GDN), whose recurrent state summarizes the context in fixed size. Early community 4-bit quantizations of Qwen3.8-27B (48 GDN layers, 16 attention layers) left the GDN block in 8- or 16-bit precision -- especially its decay and write-strength gates -- on the intuition that errors in a recurrence accumulate over long contexts. We test that intuition by building Minima: NVFP4 W4A4 on all 496 linear layers, GDN included. Across perplexity at 4K/32K, MMLU-Pro, GSM8K, AIME'25, GPQA-Diamond, LiveCodeBench, and RULER retrieval to 64K, Minima matches BF16 within seed noise (5-task average -0.52) while being the smallest (17.5 GiB) and fastest-prefill (+14-19%) recipe we compare, and its 32K perplexity gap shrinks with position. A four-part mechanism study explains why: (i) NVFP4's 16-element block scaling localizes the residual stream's extreme outliers, equalizing activation error across layer roles; (ii) the supposedly fragile gate projections are the least sensitive -- softplus/exponential and sigmoid parameterizations compress ~11% GEMM error to ~2% output error; (iii) the delta-rule recurrence holds injected noise at a flat plateau over 32K tokens and forgets a state impulse within hundreds of steps, because each write overwrites the state along the current key direction; (iv) the per-token quantization cost washes out with context instead of compounding. We also repair a global-scale mismatch that arises when per-module-calibrated NVFP4 checkpoints are served by kernels that fuse those modules into one GEMM, and show calibrated FP8 KV-cache scales are performance-free. The result: a practical recipe -- quantize everything, ship KV scales -- and a mechanistic account of why the recurrent half of a hybrid LLM is the easy half to quantize. Checkpoint: https://huggingface.co/minima-ai/mnma_qwen3.8_27b_nvfp4 --- ### Adaptive Vision-Language Grasping via Composable Foundation Priors and Generalizable Grasp Synthesis Authors: Sixu Yan, Shikang Wang, Binhua Huang, Xuanlai Tang, Guohua Fan, Fan Huang, Haoxuan Li, Yongkang Li, Yuhan Li, Bencheng Liao, Zeyu Zhang, Wenyu Liu et al. Published: 2026-09-03 Categories: cs.RO, cs.AI, cs.CV arXiv: https://arxiv.org/abs/2609.04096v1 This paper proposes AdaRoboVLG, a task-adaptive Vision-Language-Grasp (VLG) framework that supports generalizable grasp synthesis across different robotic hands. Unlike existing VLG methods that tightly couple foundation models with end-to-end grasp policies, AdaRoboVLG learns an efficient generalizable base policy that generates and evaluates physically feasible grasp candidates through explicit kinematic mapping and force-closure-based stability estimation, while offloading task-dependent understanding to specialized foundation-model modules. These modules provide composable priors that are integrated into the grasp synthesis process, enabling contextually adaptive grasp synthesis without retraining the underlying grasp policy. Through extensive simulation and real-world experiments, we demonstrate that (i) the base policy exhibits efficient learning and strong cross-hand generalization, (ii) the framework effectively incorporates spatial, cognitive, and temporal priors to address three representative grasping challenges without compromising grasp synthesis performance compared to state-of-the-art methods, and (iii) these priors can operate jointly to enable functional grasping in cluttered and dynamic environments. These results indicate that decoupling physical grasp synthesis from task-dependent understanding provides a scalable paradigm for robotic grasping, allowing future advances in foundation models to be directly translated into improved grasp capabilities without redesigning or retraining the underlying grasp policy. Supplementary videos are available at https://adarobovlg.github.io/ --- ### DRACO: Fine-Grained Credit Assignment with Dynamic Rubrics for Long-Horizon Agent Training Authors: Shubham Gandhi, Saurabh Goyal, Kiran Kate, Yara Rizk Published: 2026-09-03 Categories: cs.AI, cs.LG, cs.SE arXiv: https://arxiv.org/abs/2609.04094v1 Reinforcement Learning from Verifiable Rewards works well when a task has a programmatic checker, but most long-horizon agent domains have none. We work in the outcome-blind setting, where ground-truth success signals are not available. Multi-criteria rubrics are a popular way to supply such a reward; they are scored once per trajectory, but a single scalar is a poor signal across tens of steps. We propose DRACO: Distributing Rubric-based Advantage for Credit Optimization. It generates rubrics dynamically during training to track the policy's evolving capability, scores those rubrics once per completed trajectory, and redistributes that judgment over the steps responsible for annotated rubrics to produce differentiated per-step advantages in GRPO. The redistribution is closed-form and does not introduce any trained attribution module. On AppWorld, DRACO gains 15.9 points over the base model and 5.3 points over GRPO trained with a sparse ground-truth reward, despite not using any verifiers itself. On out-of-domain Tau-Bench, it gains 5.3 points over the base model even without a frontier judge, beating both ground-truth-reward training and other rubric-based training settings. The code for DRACO is available at https://github.com/IBM/draco. --- ### Conditioning Degenerate Diffusion Models Authors: Uğur Aydın, Tamer Başar Published: 2026-09-03 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.04090v2 Current conditioned generative models heavily rely on score functions for guidance during training. When the generative model is a diffusion process with a singular diffusion coefficient and the underlying (conditional) densities either do not exist or are not smooth, we use causal optimal transport to define \emph{approximate} loss functions that identify a minimum-entropy control for guidance under minimal assumptions. Our approach relies on causal optimal transport and its characterization through the predictable representation property of (conditioned) diffusion processes whose associated martingale problem is well posed, à la Üstünel. --- ### Conditioning Degenerate Diffusion Models Authors: Uğur Aydın, Tamer Başar Published: 2026-09-03 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.04090v1 Current conditioned generative models heavily rely on score functions for guidance during training. When the generative model is a diffusion process with a singular diffusion coefficient and the underlying (conditional) densities either do not exist or are not smooth, we use causal optimal transport to define \emph{approximate} loss functions that identify a minimum-entropy control for guidance under minimal assumptions. Our approach relies on causal optimal transport and its characterization through the predictable representation property of (conditioned) diffusion processes whose associated martingale problem is well posed, à la Üstünel. --- ### Efficient Semantic Understanding from Digital Foveation Authors: Caterina Caccavella, Vittorio Fra, Andreas Ziegler, Giulia D'Angelo, Yulia Sandamirskaya Published: 2026-09-03 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.04088v1 Dense semantic segmentation allocates computational resources uniformly across the entire image, regardless of scene complexity or task relevance. Inspired by biological vision, we investigate whether semantic understanding can be achieved more efficiently through digital foveated perception. We introduce a lightweight active-vision pipeline that combines saliency-driven fixation selection, high-resolution foveal observations, low-resolution contextual information, semantic accumulation, and adaptive computation. Beyond conventional dense prediction metrics, we use object-level evaluation to measure semantic understanding under sparse observations. On ADE20K-Object, a single foveated observation achieves 95.9% of the baseline Top-1 accuracy and 96.9% of the baseline Top-3 accuracy while requiring only 4.7% of the computational cost. At the scene level, semantic accumulation recovers 90.6% of the baseline object recall while using 58.6% of the computation. These results suggest that substantial semantic understanding can emerge from sparse observations when computation is allocated selectively, highlighting active vision as an efficient alternative to uniform dense processing and motivating evaluation protocols beyond conventional pixel-wise segmentation metrics. --- ### A Non-Formulable Theorem: A Fundamental Limit of Finite Syntactic Systems and Its Consequences for Security and AI Authors: Fabio F. G. Buono Published: 2026-09-03 Categories: cs.CR, cs.AI, cs.LO arXiv: https://arxiv.org/abs/2609.04086v1 For every coherent and sufficiently expressive finite syntactic system S, we prove the existence of at least one theorem that S cannot produce autonomously. The result is a metatheorem: it proves the existence of a theorem, and applies to every finite syntactic system - security mechanisms, AI systems, formal verifiers, legal systems, economic models, and the formal system in which it is itself proved. --- ### CORE: Improving Compositional Reasoning in MLLM Embedding via Reranker Distillation Authors: Tingyu Song, Mingxin Li, Yanzhao Zhang, Dingkun Long, Chu Liu, Pengjun Xie, Yilun Zhao, Shu Wu Published: 2026-09-03 Categories: cs.CV, cs.AI, cs.CL, cs.IR arXiv: https://arxiv.org/abs/2609.04083v1 MLLM-based embedding models remain limited in compositional retrieval, often failing to distinguish scenes containing the same concepts but different attribute-object bindings. Yet the same backbone can resolve such distinctions when used as a cross-attentive reranker, motivating us to distill its compositional judgments into the embedding model. We propose CORE, which synthesizes candidate lists spanning five compositional matching levels and introduces a Rank-KL objective that trains the embedding model to reproduce the reranker's fine-grained ranking. We further introduce a graded evaluation protocol and compare contrastive learning, pairwise CoSENT, and listwise Rank-KL under the same data and tuning budget. Our comparison shows that both CoSENT and Rank-KL use the multi-level supervision more effectively than contrastive learning, with Rank-KL achieving the strongest overall performance. Across three compositional reasoning benchmarks (COLA, SUGARCREPE++, NEGBENCH), CORE-RERANKER-8B achieves an 82.7% total average, outperforming Jina-Reranker by 10.7 points, while CORE-EMBED-8B achieves the best total average (0.666) among all evaluated embedding models. The improvements transfer to the MCMR benchmark without sacrificing retrieval performance on COCO and Flickr30K. ---