Knowledge
-
Memorilla: Latent Semantic Memory for LLMs
2026-10-15 06:55:55Adibvafa Fallahpour, Parsa Idehpour, Vignesh Kothapalli, Nikita Mounier, Shirley Wu, Shayan Pardis, Jure LeskovecAbstract
Memory enables large language models (LLMs) to use information beyond their parameters, such as past user conversations, long documents, or an agent's earlier actions. For a frozen LLM, access to this accumulated knowledge is limited to what a memory system places in its context window. However, gathering and combining relevant information needed for each query while keeping the model's context short remains challenging for existing systems. Here we introduce Memorilla, a memory system that gathers information from an entire document collection into a fixed number of query-dependent 'memory tokens' for a frozen LLM. Memorilla encodes documents (such as user chats, book chapters) once into reusable embeddings, then learns to retrieve and place query-specific information in the LLM's input embedding space. Using only 16 memory tokens, Memorilla achieves the best performance against baselines including Mem0 and MemoRAG on five of eight benchmarks spanning personalization, long-document and knowledge-intensive question answering. Notably, on FactKG, it reaches 84.33% accuracy versus 80.43% with full-context inference, while using 108x fewer tokens to represent the source documents. Furthermore, Memorilla can facilitate continual learning in agentic systems where the document count (the agent's past turns) increases throughout the trajectory. In a TextWorld environment, we show that Memorilla reduces the agent's tendency to wander between rooms it has already explored, and improves its average success rate from 18.6% with full-history in-context learning to 24.2%. Overall, Memorilla aids frozen LLMs in accumulating and using knowledge effectively across long-context interactions and takes a step towards building continually learning AI systems.
Abstract (translated)
URL
https://arxiv.org/abs/2610.21375
PDF
-
RenoLayout: Unified Floorplan Renovation and Furniture Layout for Constraint-Aware Residential Design
2026-10-15 06:44:30Jingguo Liu, Yuanqing Liu, Chong Zhang, Yue YangAbstract
Residential design aims to transform existing homes into functional living spaces while satisfying diverse design preferences and physical constraints. However, existing floorplan generation methods typically assume empty building boundaries as input, whereas renovation requires reasoning over existing floorplans. Furthermore, floorplan renovation and furniture layout are commonly addressed as separate tasks, overlooking their inherent spatial dependency. Consequently, practical residential design must preserve immutable structural components while transforming editable spatial elements. It must also optimize furniture placement under design objectives and physical constraints. To address these challenges, we formulate residential design as a design-point-guided conditional generation task and propose RenoLayout, a framework that translates a pre-design residential state together with a design-point set into an executable Geometry-as-Code program. To enable controllable supervision, we construct a Design Map comprising 123 fine-grained design points covering structural renovation and furniture layout. We further introduce Difference-Grounded Design-Point Annotation to derive design-point sets automatically from before--after design pairs with VLM-based consistency verification. RenoLayout is trained using a three-stage CPT--SFT--RL pipeline, progressively improving spatial reasoning, instruction following, and constraint compliance. We also introduce the RenoLayout Benchmark, a set of 200 real-world before--after renovation cases with complete furniture layouts. To the best of our knowledge, it is the first benchmark for unified residential design. These cases include design-critical structural and building-service constraints.
Abstract (translated)
URL
https://arxiv.org/abs/2610.21366
PDF
-
FedSkillGit: Federated Skill Evolution via Version Control
2026-10-15 06:29:57Boxun Lan, Jiamei Li, Weijian Li, Chengyu Yang, Zhuang Qi, Chengcheng Wan, Xiaohong Chen, Ming HuAbstract
Large language model agents can improve their capabilities by evolving reusable natural-language skills from execution experience. In real-world deployments, this experience is distributed across data owners who cannot pool their raw data, which motivates federated skill evolution. The natural first design lets each client evolve the whole skill document and upload the result for server-side aggregation. That design is poorly matched to skill evolution for three reasons. An evolved document can still carry client-specific content distilled from private traces, repeated transmission of a nearly unchanged document spends communication on knowledge that the population already holds and the contribution of a single client is buried in a document that the server cannot meaningfully attribute. We propose FedSkillGit, a federated skill evolution framework that replaces full-document updates with commits. Each participating client submits a bounded set of patches over named rules together with its local experience, which summarizes what the round taught it, and the pair passes a two-stage privacy filter before it leaves the client. The server deduplicates the submitted patches and the local experience of all participants, updates a global experience document that only the optimizers read, and merges the surviving patches into a candidate skill. The candidate is distributed to the clients, which score it on their own validation sets and report the results, and it is published only when the reported scores improve on the current version. Experiments on three datasets and three language model families show that FedSkillGit outperforms state-of-the-art baselines by up to 22% and approaches centralized evolution without sharing raw client data.
Abstract (translated)
URL
https://arxiv.org/abs/2610.21348
PDF
-
Emotion Is Not External Knowledge: Internal Attention Signals for Affective State Reasoning
2026-10-15 06:29:55Wenqiu Tang, Zhen Wan, Yihui Wang, Takahiro Komamizu, Ichiro IdeAbstract
Large Language Models (LLMs) increasingly operate in multi-turn interactions that require maintaining latent conversational states, not only explicit facts. We study affective state as an implicit state inferred from sparse contextual cues that remain relevant after topic shifts, and ask whether LLMs expose internal attention signals for routing such evidence at inference time. Prior attention interventions rely on labeled in-domain head profiles and externally specified target spans, limiting transfer across affective-evidence distributions. We propose Self-targeted Internal Attention Boosting (SIAB), which localizes sparse affective heads via causal localization (in detail activation patching) on synthetic minimal pairs and boosts, at the answer position, the content tokens these heads already attend to; both edited heads and target tokens are derived from the model's own behavior, enabling the same head set to be reused across this http URL evaluate on ECC and GoEmotions across three LLMs and multiple prompting settings. Across settings, SIAB improves prompt-matched baselines by up to +10.3 macro-F1 on ECC and +5.5 on GoEmotions. Combined with PASTA-style external supervision, it achieves up to +13.84 accuracy points and +12.51 macro-F1 points, while mechanistic analyses trace the gains to sparse, emotion-enriched heads and show that layer proximity affects additivity.
Abstract (translated)
URL
https://arxiv.org/abs/2610.21347
PDF
-
OpCredit: Operation-Aware Policy Learning for Agentic Knowledge Graph Question Answering
2026-10-15 06:29:17Yucheng Li, Mingyuan Li, Xiaxia Wang, Gong ChengAbstract
Agentic knowledge graph question answering(KGQA) relies on sequences of executable operations, but trajectory-level rewards obscure their individual contributions. Successful trajectories may contain ineffective operations, while failed trajectories may contain useful ones. To assess individual operations, we propose OpCredit, an operationaware policy-learning framework that combines downstream outcome credit with immediate transition credit: Branch Outcome Credit (BOC) reuses ordinary rollouts to compare terminal task rewards observed after different operations under matched navigation contexts; Frontier Transition Credit (FTC) measures operation-induced changes in answer-conditioned frontier potential. The two complementary signals jointly supervise the corresponding decision spans during training, while inference uses only the learned policy and standard KG interface. Experiments on ComplexWebQuestions, WebQSP, and GrailQA across three backbones demonstrate consistent improvements in KGQA performance.
Abstract (translated)
URL
https://arxiv.org/abs/2610.21345
PDF
-
ZoomCap: Evidence-Routed Visual Zooming for Recoverable Visual Information Transfer
2026-10-15 06:10:24Hao Shi, Jiafeng Wang, Yuzhuo Fu, Tao Zhang, Jian Bai, Juelu Zhang, Qing Cui, Chao HuangAbstract
Detailed captions transfer visual information into vision-language model supervision, but quality is easily conflated with length: long single-pass outputs still miss small objects, text, and spatial relations or promote uncertain observations into confident claims. We formulate fine-grained captioning as recoverable visual information transfer, in which a teacher acquires missing evidence under a finite observation budget and converts it into reusable supervision. ZoomCap uses a draft caption to route conditional knowledge retrieval, OCR-sensitive reading, coverage-gap recovery, and adaptive visual zooming, allocating deeper crops only to the most complex eligible region, deriving spatial relations from grounded geometry, and fusing non-duplicate evidence. We also introduce FactCycle, which scores caption-derived atomic claims on two blind anchors--source image (truthfulness) and reconstruction (realizability)--separating fabrication from generator loss claim by claim; human and multi-judge comparisons complement this fact-level view. Against Qwen3.5-397B-A17B captions, ZoomCap achieves an 81.9% human tie-adjusted preference score with a teacher over 10x smaller, and FactCycle counts 52.4 transferred claims per image vs. 31.5 for ScaleCap. The evidence also transfers: in continual pre-training, ZoomCap-450K improves the 11-benchmark average by 1.31-2.38 points over the strongest competing corpus and Ling-Tiny-VL-v3 gains 1.4 points on a production document/OCR monitor suite (1.5 on perception) at 60B tokens; in medical pre-training, 2.64M ZoomCap-captioned images improve five VQA benchmarks by 3.04 points (vs. 1.76 for same-image Gemini-3.1-Pro captions).
Abstract (translated)
URL
https://arxiv.org/abs/2610.21320
PDF
-
VLA-$Delta$: Learning Physical State Transition for Vision-Language-Action Models
2026-10-15 05:54:21Qi Lu, Jiarun Li, Ce Zhang, Miao Pan, Yi Pan, Xu Wang, Wenqi Zhang, Xuhong ZhangAbstract
Vision-language-action (VLA) models transfer semantic knowledge acquired through vision-language pretraining to robot control, achieving strong performance across various benchmarks. However, most VLAs directly map observations to demonstrated actions without explicit supervision of intermediate physical state changes, making them prone to overfitting observation-action correspondences. Some methods introduce intermediate reasoning or future-state prediction in language, visual, or latent spaces to guide action generation, but do not explicitly supervise the physical state transition itself. We therefore propose VLA-$\Delta$, a general framework that uses physical changes between current and future states as explicit learning targets. To support this objective, we construct a large-scale pretraining dataset and derive physical state-transition annotations covering task-relevant semantics and visual motion changes, providing more direct supervision for continuous action generation. During inference, the model generates actions directly from current observations without additional inference overhead. VLA-$\Delta$ achieves performance competitive with the state of the art on LIBERO and RoboTwin, and outperforms $\pi_{0.5}$ in real-robot experiments. Our code is available at this https URL
Abstract (translated)
URL
https://arxiv.org/abs/2610.21306
PDF
-
Bias Control and Injection for Multi-Agent Aggregation
2026-10-15 05:37:49Sungjae Lee, Soheil Abbasloo, Jabin Koo, Jungseul OkAbstract
Multi-agent reasoning with large language models (LLMs) raises a central question for collective decision-making: when agents disagree, whose decision should be trusted? Resolving such disagreements is difficult because agents' decisions are susceptible to non-semantic biases such as majority pressure and response-style preferences, which can obscure the quality of the underlying reasoning. Our analysis shows how bias can be both mitigated and exploited: bias control improves initial answer selection, while bias injection elicits decision instability that serves as a reliability signal. Building on these findings, we introduce a test-time aggregation framework that requires neither validation data nor prior knowledge of relative model strength. The framework implements bias control by presenting one style-normalized reasoning trace per distinct answer, and bias injection by placing each aggregating agent's initially selected answer in the minority. Overall, our framework achieves higher average accuracy and better token efficiency than competing non-oracle aggregation methods across three mathematical reasoning benchmarks and four heterogeneous LLM combinations. It also improves final-answer accuracy when applied to multi-agent debate and Mixture-of-Agents, without modifying their interaction or synthesis processes.
Abstract (translated)
URL
https://arxiv.org/abs/2610.21290
PDF
-
Diviner: Agentic RCA with Human-Governed Adaptation for Large-Scale Infrastructure
2026-10-15 05:20:23Pradipta Ghosh, Gaurav Mitra, Shripad Nadgowda, Akshil Shah, Junyu Liu, Canturk Isci, Syamla Bandla, Xinyao Hu, Onkar Deshpande, Anik DasguptaAbstract
Root-cause analysis (RCA) in large-scale distributed systems is challenging because a single symptom can cross service, data, configuration, compute, and organizational boundaries. Responders must assemble incomplete evidence and coordinate experts who hold different parts of the causal path. Generic language-model agents do not solve this problem through tool access alone: they can use stale or mistimed context, interpret failed queries as absent symptoms, report unsupported system state, and overload observability services designed for human-paced use. We present Diviner (Domain-Informed Verifiable Incident Navigation through Evidence and Review), an agentic RCA system operating at Meta and organized around a four-stage reliability lifecycle. Encode combines a large corpus of reviewed incidents with service-owner expertise to produce scoped, versioned skills, workflows, typed tools, and an incident knowledge graph. Execute uses an adaptive Lead and context-isolated workers to apply relevant knowledge to point-in-time evidence and maintain evidence-linked hypotheses. Evaluate compares saved investigation trajectories with responder-reviewed outcomes to measure exact cause, useful direction, and Groundedness. Adapt uses recurring gaps and on-call feedback to derive focused operational- knowledge updates, admitting them only after replay, quality checks, and owner approval.
Abstract (translated)
URL
https://arxiv.org/abs/2610.21281
PDF
-
System1-Arena: Which Decisions Can Be Left to System 1?
2026-10-15 04:40:47Tianyu Chen, Xinyang Liu, Ruyi An, Zhendong Wang, Mingyuan ZhouAbstract
Large language models (LLMs) often make discrete decisions, such as flagging phishing or escalating to a human. System-1 models, such as Jev and the OpenAI Decisions API, return probabilities over options in a single forward pass, while System-2 approaches generate reasoning steps. Which decisions can System-1 models handle, and which require generative reasoning? We introduce System1-Arena, a benchmark that gives all models the same decision requests and scores their probabilities. It contains 92k questions across 105 tasks from 268 public sources, in four groups: text classification and knowledge, applied decisions, evaluation and control, and agent and action. It also includes 138 closed-loop environments, from web shopping and SWE-bench coding control to robotics, games, and driving; in real-time mode, the world evolves while a model decides. We evaluate 30 models: 7 closed LLMs, 13 open LLMs read at the answer token, and 10 System-1 models. The largest System-1 models match or exceed fast closed LLMs in non-agentic accuracy, with better-calibrated probabilities at one-quarter of the latency. A reasoning LLM is the most accurate, at about 70x the cost of a fast LLM. In closed loops, large open LLMs succeed most often (up to 32.9%); small System-1 models rarely succeed.
Abstract (translated)
URL
https://arxiv.org/abs/2610.21244
PDF
-
An Agentic Framework for Proactive Robots with Procedural Assistance Memory
2026-10-15 04:04:49Dohyun Lee, Minho Park, Jeonghoon Park, Dongyoon Hwang, Junha Hyung, Sungwon Hwang, Byungkun Lee, Jaegul ChooAbstract
Proactive robots should anticipate user needs and provide assistance before being explicitly asked. A straightforward approach is to infer human intentions from commonsense knowledge, but the same observed behavior can imply different needs across users, potentially leading to unwanted assistance. We observe that everyday routines in the real world exhibit recurring patterns, so learning these patterns can reduce ambiguity in proactive decisions. Based on this, we propose a memory-based agentic framework in which a VLM task predictor reads a Procedural Assistance Memory (PAM) alongside the current observation to decide when to assist and what task to perform, and a VLA executor performs the predicted task. PAM is structured as a finite-state machine over activity phases, per-state assistance rules, and event-driven transitions; a reflection VLM revises PAM from each day's interaction trace without weight updates or retraining. In our 5-day desk-assistance evaluation, memory-free baselines remain conservative, with limited recall on proactive tasks. Our system improves consistently with accumulated experience, achieving the highest precision, recall, and F1 and reducing first-fire delay from approximately 5 seconds on Day 1 to 1 second by Day 5. Our approach also adapts to changes in the user's routine through continued memory reflection, recovering from memory--routine misalignment after a single memory update. The project page is available at this https URL
Abstract (translated)
URL
https://arxiv.org/abs/2610.21211
PDF
-
LIN-SLAM: Lightweight LiDAR--Inertial Neural SLAM for Real-Time Onboard Navigation
2026-10-15 03:58:47Shiqi Che, Yijie Liu, Shuo Wang, Wenxin LiuAbstract
Geometric light detection and ranging (LiDAR)--inertial simultaneous localization and mapping (SLAM) provides efficient state estimation but lacks the continuous map representation and global refinement capabilities of neural implicit SLAM, while neural implicit methods remain computationally expensive. We propose Lightweight LiDAR--Inertial Neural SLAM (LIN-SLAM), a hybrid method that couples high-rate geometric LiDAR--inertial estimation with neural implicit tracking, continual mapping, and global refinement for accurate real-time onboard SLAM. Experiments on twelve representative sequences from three datasets spanning 32-, 64-, and 128-channel LiDAR configurations show that LIN-SLAM achieves the lowest overall mean absolute trajectory error (ATE) among the evaluated state-of-the-art geometric and neural implicit methods, while substantially reducing runtime compared with existing neural implicit SLAM methods. On a Unitree Go2 equipped with an NVIDIA Jetson Orin NX and a 16-channel LiDAR, LIN-SLAM sustains real-time localization and neural mapping over a 1.47-km indoor trajectory through challenging repetitive corridor environments, with an endpoint closure error of only 0.141~m. To the best of our knowledge, LIN-SLAM is the first LiDAR--inertial SLAM system with a neural implicit backend to achieve real-time online operation on a mobile robotic platform while maintaining state-of-the-art localization accuracy. The code will be released upon acceptance.
Abstract (translated)
URL
https://arxiv.org/abs/2610.21199
PDF
-
ControlFlow3D: Distilling Multi-View Knowledge into Latent Flow Matching for Point Cloud Upsampling
2026-10-15 03:50:42Yuang Liu, Zhi Zuo, Zhengkai Zhao, Lirui Zhang, Pan Gao, Hao Feng, Zhengzhe LiuAbstract
Point cloud upsampling is essential for 3D applications but remains challenging due to sparse, under-constrained observations. Multi-view images can resolve geometric ambiguity, but existing multi-modal methods do not fully exploit visual cues. We propose ControlFlow3D, a framework that distills knowledge from pretrained foundation models into a lightweight generation network during training, enabling strong performance without multi-view images available at test time. Specifically, our method performs conditional flow matching in a structured latent space defined by a frozen 3D encoder. Multi-view image tokens and 3D latent tokens are jointly clustered into shared prototypes via unsupervised soft KMeans; the resulting prototypes guide the flow trajectory through velocity control and are aligned by a consistency loss that progressively transfers image knowledge into the base network. A Mamba-based geometry decoder then maps latent features to dense coordinates with linear complexity. We also construct ShapeNetPU, a 35K-object multi-modal benchmark that is 31 times larger than prior datasets. Extensive experiments across multiple benchmarks demonstrate the effectiveness of our method and its strong generalization capability.
Abstract (translated)
URL
https://arxiv.org/abs/2610.21190
PDF
-
Evo-DexBench: Benchmarking Generalization across Related Tasks for Dexterous Manipulation
2026-10-15 03:36:05Bing Cheng, Yiran Mao, Zewei Ye, Yifan Jin, Xinyi Chen, Ruifeng Tu, Shuhuai Wang, Tao Lin, Zhimu Zhou, Yuxin Du, Fangwei Zhong, Junchi Yan, Bo ZhaoAbstract
General-purpose dexterous manipulation requires policies to transfer knowledge from related tasks to new tasks with only a few target demonstrations. Existing benchmarks often provide too many target demonstrations. They also leave the differences between source and target tasks uncontrolled. As a result, they cannot clearly distinguish knowledge transfer from task memorization. Our key insight is the Core-Gen principle. Core tasks are related tasks that share a functional interaction and provide source knowledge, while a Gen task is a novel target within the same functional family that has only a few demonstrations. This setup tests whether a policy can reuse functional interaction knowledge. Based on this principle, we propose Evo-DexBench, which controls both source-task relatedness and target-task exposure. It comprises 30 tasks in six functional families and a VR teleoperation pipeline. The resulting dataset covers three everyday scenes per task. It contains 1,248 multimodal demonstrations collected with simulated Shadow Hands equipped with 123 taxels. Experiments show that even the best-performing method achieves an average success rate of only 38.5%, highlighting the challenges existing methods face in transferring functional interaction knowledge. Further analysis shows that, across the two tasks, joint Core+Gen training improves performance by 12% and 38%, respectively, over training on Core followed by fine-tuning on Gen. This indicates that joint training makes better use of Core knowledge and that functionally matched Core data yields greater gains in success rate than mismatched data. We release the Evo-DexBench dataset and benchmark to support the evaluation of dexterous manipulation policies' generalization across related tasks.
Abstract (translated)
URL
https://arxiv.org/abs/2610.21180
PDF
-
DST-Net: Decomposed Spatio-Temporal Network for Time Series Forecasting
2026-10-15 02:28:22Yaohua Zha, Hai Wu, Shu-Tao XiaAbstract
Series decomposition has emerged as a widely acknowledged paradigm in deep time series forecasting models. However, existing approaches predominantly employ a homogeneous modeling strategy for all subsequences, thereby overlooking the distinct inductive biases inherent in different components. We identify that the trend component exhibits dominant spatial coupling across channels but carries significant temporal redundancy, whereas the seasonal component is characterized by local temporal evolution but remains susceptible to attention distraction. To address this, we propose a novel disentangled architecture that tailors specific modeling strategies to align with the intrinsic properties of each component. For the trend branch, we design a Latent Graph Block, employing latent graph modeling to capture global topological dependencies and system-wide co-movements. Conversely, for the seasonal branch, we introduce a Gaussian Multi-scale Attention mechanism to mitigate attention distraction. By injecting learnable Gaussian priors, we impose structural constraints that compel the model to focus on multi-granular local contexts, effectively filtering out global noise and accurately capturing local evolutionary phases. Extensive experiments on multiple real-world benchmarks demonstrate the effectiveness of our proposed model.
Abstract (translated)
URL
https://arxiv.org/abs/2610.21123
PDF
-
Planning Commitment of LLM under Partial Observability
2026-10-15 02:11:28Cassie Huang, Zhanwei Cao, Yeawon Lee, Nathalia Gomez, Zhuoyi Yang, Ian G. Harris, Tiffany Do, Jiayuan Mao, Li ZhangAbstract
LLM-based planning agents commit to plans of different lengths before interacting with a partially observable environment. Some aggressively commit to a long plan based on assumptions about unobserved states, while others conservatively plan only over what they have observed. Both behaviors emerge implicitly in pre-trained models, whether the LLM outputs a plan directly or generates a PDDL formulation for a classical solver. Aggressive commitment risks costly failures, while conservative commitment risks excessive LLM calls. This choice pits safety against efficiency, a trade-off that has been neither characterized systematically nor made controllable. We investigate aggressive and conservative LLM planners and formalizers on two partially observable domains, where priors are provided as either explicit statistics or common-sense knowledge. Across three LLMs, both strategies reach similar accuracy. Conservative methods achieve fewer failed attempts (safety) at the cost of more API calls (efficiency). Aggressive methods fail more often but replan efficiently only when priors are informative. Interestingly, analysis shows that LLMs treat the provided priors nearly as facts, which risks failure when priors are unreliable. We also explore hybrid strategies with tunable hyperparameters that adjust the balance between safety and efficiency.
Abstract (translated)
URL
https://arxiv.org/abs/2610.21108
PDF
-
Athanor: An Agent-Native System for Autonomous Post-Training Research
2026-10-15 00:56:33Yufa Zhou, Yichen Wang, Jiakun Fan, Ziqi Shu, Zhichao Liu, Anru R. Zhang, Neiwen Ling, Zijun WeiAbstract
Recursive self-improvement, in which AI systems build stronger AI systems, is one promising route to superintelligence. We study a weaker form that can be measured today: frontier agents post-training a small raw base model with public data, as a test of their research potential. We build Athanor, an open agent-native system for autonomous post-training research (AutoPostTrain) on real GPUs. We give the agents full research autonomy, low-friction engineering tools, and efficient starter code. The agents carry out the research themselves: they read datasets and evaluation results, build their training data, write and modify training algorithms, tune hyperparameters, and analyze the results to decide what to try next. They work in a sandbox and cannot use external teacher models. Athanor's services supply frozen evaluation, managed jobs, decontaminated data, a persistent research record, and a held-out suite the agents are unaware of. In a 104-hour campaign, a team of four Claude Opus 5.5 agents, each on an eight-GPU node and given no research advice, improves Qwen3-4B-Base from 15.9 to 67.3 on a six-task development suite and from 9.2 to 48.2 on a held-out suite, against 66.5 and 47.9 for the official Qwen3-4B-Instruct-2507. To our knowledge, this is the first demonstration that a team of frontier agents can autonomously post-train a raw base model into a cross-domain generalist model that matches or surpasses the official instruct version on average. The campaign's full record shows that the agents reach this level by recombining known methods and each other's models: SFT on public reasoning data, RL with a rollout change one agent introduced, weight merges across agents, and rounds of self-distillation. With execution handled, today's agents already have the engineering and research ability to conquer this task. We hope Athanor provides a foundation for scaling autonomous AI research.
Abstract (translated)
URL
https://arxiv.org/abs/2610.21055
PDF
-
INCONVENIENT: A Benchmark to Assess LLM Generalization Capabilities from Convenience Samples
2026-10-15 00:15:12Sara Rezaeimanesh, Chad Klochko, Kundan Thind, Mohammad GhassemiAbstract
Machine learning models for tabular prediction typically assume that training data are representative of the population on which they will be deployed. In practice, however, data are often collected based on convenience, availability, or cost, resulting in systematic sampling biases that can distort feature--target relationships and undermine real-world generalization. Existing tabular distribution-shift benchmarks primarily construct out-of-distribution (OOD) settings by partitioning observations across naturally occurring domains. While such partitions may induce changes in P(Y | X), the resulting shifts are neither explicitly controlled nor isolated, and the evaluation distributions themselves are often not representative of the broader target population. We introduce INCONVENIENT, a benchmark designed to study robustness to convenience-sample bias. Rather than relying on naturally occurring domain partitions, INCONVENIENT directly perturbs feature--target relationships in the training distribution while preserving the original population distribution as the deployment environment. This setting more closely reflects real-world scenarios in which models are developed from unrepresentative samples and subsequently deployed on a broader population. The distinction is particularly relevant for large language models (LLMs), whose pretraining may encode population-level knowledge that extends beyond the relationships observed in a biased training dataset. Unlike existing distribution-shift benchmarks, INCONVENIENT provides explicit control over both the direction and severity of relationship shifts, enabling a systematic assessment of robustness to convenience sampling. The benchmark comprises 200 tabular classification and regression datasets spanning diverse application domains and generates around 15,500 shifted training distributions which lead to substantial downstream degradation.
Abstract (translated)
URL
https://arxiv.org/abs/2610.21038
PDF
-
MoCaP: Mobile Code-as-Policies for Zero-Shot Mobile Manipulation from Passive Human Videos
2026-10-14 22:33:20Yiyang Cai, Tatsuya Kamijo, Ninghao Lu, Zijin Hu, Sherry YangAbstract
Passive human videos of everyday activity offer robots access to a massive amount of human experience without requiring a demonstration for each task. However, using this experience requires recovering knowledge from unstructured video recordings and adapting it to a different body (i.e., the robot). To overcome these challenges, we present mobile code-as-policy (MoCaP), an agentic framework that turns passive human video recordings into guidance for an end-to-end mobile manipulation policy. MoCaP extracts spatial memory and cues of how people prepare to interact with objects, which inform navigation and the search for a manipulation-ready configuration. Guided by passive human experience and current observations, MoCaP uses a closed-loop code-as-policy to coordinate planning and tool use primitives to solve motions under the robot's geometry and kinematic constraints. On a real mobile manipulator, MoCaP achieves 80% success on each of two mobile manipulation tasks in an indoor environment whereas existing approaches cannot succeed at the task. Video-derived spatial memory improves object-navigation success from 70% to 100%, and human approach prior raises the rate of finding a reachable base pose from 0% to 100% at an unfavorable start location. Videos and code are available on the project website: this https URL.
Abstract (translated)
URL
https://arxiv.org/abs/2610.20998
PDF
-
Do As I Do, Not As I Say: On-Policy Distillation Inherits the Teacher in Vision-Language Models
2026-10-14 22:04:32Frank F. Yang, Kefei Yan, Yiren Wang, Yiyang WangAbstract
Improving the reasoning capabilities of vision-language models (VLMs) often relies on post-training with supervision from stronger models. Two common strategies are supervised fine-tuning (SFT), which trains on teacher-generated demonstrations, and on-policy distillation (OPD), which provides teacher feedback on trajectories generated by the student itself. While OPD can be more data-efficient by supervising student-generated trajectories, it is unclear how the teacher's supervision at each state affects the student's learned behavior. To study this question, we conduct a controlled comparison of OPD and filtered SFT across data budgets and tasks where the teacher exhibits different levels of reliability, and analyze how these differences shape student accuracy and completion behavior. Across the two tasks, we find that OPD is effective only when the teacher remains reliable on student-generated trajectories. On ChartQA, where the teacher completes 99% of its solutions, OPD trained on only 10% of the data comes within 1.4 accuracy points of full-data SFT. In contrast, on Geometry3K, where the teacher fails to complete 22% of its solutions, OPD underperforms SFT by 7.9 points, with most of the gap explained by incomplete responses. More strikingly, starting OPD from an SFT student reduces its completion rate from 76% to 62%, effectively erasing the completion behavior learned through SFT. Filtering teacher demonstrations for completeness alone recovers much of SFT's advantage. These results show that on-policy supervision can transfer not only teacher knowledge but also persistent failure modes on student-visited states. Code is available at this https URL.
Abstract (translated)
URL
https://arxiv.org/abs/2610.20989
PDF
