Open-HumanScale: An Open Recipe for Scaling Human and Robot Experience for Embodied Foundation Models
Juncheng Ma, Jiaxin Li, Jiankai Tu, Bo Liang, Guodong Lv, Yufan Deng, Junzhe Tang, Liangzhu Yang 等 31 位作者
AI 速览：提出首个全开放的第一人称人类视频预训练扩展方案，构建5万小时带标注语料并探索人类—机器人联合预训练的世界—动作模型。
Egocentric human video offers a scalable source of manipulation priors for robot learning, yet translating raw data scale into generalizable robot policies demands rigorous data curation and training recipes. We introduce Open-HumanScale, the first fully open recipe for scaling egocentric pretraining to 50,000 hours, anchored by Open-HumanScale-50K---a curated egocentric video corpus densely annotated with hand motion and natural language. Leveraging this corpus, we systematically investigate pure human pretraining via closed-loop simulation evaluations and uncover an empirical log-linear scaling law linking pretraining data hours to downstream robot success rates. Grounded in these findings, we develop HumanWAM-Vanilla, a world--action model pretrained exclusively on 50k hours of egocentric human video. Upon robot post-training, HumanWAM-Vanilla reaches 92.08% success on RoboTwin 2.0 Clean2Clean and 68.88% on Clean2Random, alongside a 14.17% mean success rate and a 19.00 mean progress score on RoboDojo. To push the capability frontier further, we explore joint human--robot pretraining with HumanWAM-Goal, a visual-goal-conditioned world--action model. Hybrid data scaling unlocks broad zero-shot generalization across both simulation and physical benchmarks, propelling post-trained RoboTwin 2.0 performance to new state-of-the-art levels: 92.70% on Clean2Clean and a substantial leap to 84.60% on Clean2Random. On physical robots, HumanWAM-Goal further exhibits emergent embodied in-context learning, dynamically adapting its manipulation behavior to unseen visual goals without parameter updates. To support reproducible scaling research in embodied AI, we open-source the Open-HumanScale-50K corpus, data curation and pretraining pipelines, alongside the pretrained and post-trained HumanWAM checkpoints, inference code, and simulation evaluation suites.
Fine-grained visual reasoning is essential for remote sensing in urban management, yet remains underexplored in the evaluation of multimodal large language models (MLLMs). We introduce UMBench, a fine-grained urban remote sensing benchmark built from unmanned aerial vehicle (UAV) images for multiple urban management tasks such as traffic monitoring, urban construction, environmental quality, and disaster warning. Evaluations on UMBench reveal that current MLLMs often rely on scene-level priors and overlook fine-grained visual evidence. To address this issue, we further propose a novel evidence-first chain-of-thought method that directs models to inspect fine-grained visual evidence in a training-free manner. Experiments show that our method can improve the performance of both open-source and commercial MLLMs on UMBench.
Tianwei Zhang, Ivy Njoki Nyagwencha, Zahraa Dirani, Shalini Joshi, Scott A. Hale
Tech-facilitated gender based violence (TFGBV) is a pervasive online harm affecting women and gender minority groups through its many direct and covert forms. Current models and annotated datasets for gendered harm detection are dominated by data from, and models trained in, Global North languages. In collaboration with civil society organizations (CSOs) from Jordan, Francophone West Africa, and Kenya, we developed a community-led approach to crowdsource culturally and linguistically relevant instances of TFGBV in Levantine Arabic, African French, and Swahili. We annotated the crowdsourced data using active learning to develop a small but performant dataset (N=3022) called ALTO to improve TFGBV detection in the three languages. Through benchmarking and fine-tuning experiments, we show that existing harmful content detection models show poor TFGBV recall within ALTO and improve with fine-tuning. We found that models often miss covert harms across all languages, while error rates are markedly higher in Swahili. We report the experimental results and release the annotation taxonomy, dataset, and trained models.
Gen Li, Kaifeng Zhao, Siwei Zhang, Xiaozhong Lyu, Mihai Dusmanu, Yan Zhang, Marc Pollefeys, Siyu Tang
Understanding the world in first-person view is fundamental in Augmented Reality (AR). This immersive perspective brings dramatic visual changes and unique challenges compared to third-person views. Synthetic data has empowered third-person-view vision models, but its application to embodied egocentric perception tasks remains largely unexplored. A critical challenge lies in simulating natural human movements and behaviors that effectively steer the embodied cameras to capture a faithful egocentric representation of the 3D world. To address this challenge, we introduce EgoGen, a new synthetic data generator that can produce accurate and rich ground-truth training data for egocentric perception tasks. At the heart of EgoGen is a novel human motion synthesis model that directly leverages egocentric visual inputs of a virtual human to sense the 3D environment. Combined with collision-avoiding motion primitives and a two-stage reinforcement learning approach, our motion synthesis model offers a closed-loop solution where the embodied perception and movement of the virtual human are seamlessly coupled. Compared to previous works, our model eliminates the need for a pre-defined global path, and is directly applicable to dynamic environments. Combined with our easy-to-use and scalable data generation pipeline, we demonstrate EgoGen's efficacy in three tasks: mapping and localization for head-mounted cameras, egocentric camera tracking, and human mesh recovery from egocentric views. EgoGen will be fully open-sourced, offering a practical solution for creating realistic egocentric training data and aiming to serve as a useful tool for egocentric computer vision research. Refer to our project page: https://ego-gen.github.io/.
While single-image 3D texture generation has advanced significantly, conditioning texture synthesis on multi-view reference images remains largely under-explored, despite its necessity for resolving surface ambiguity in unobserved regions. In this paper, we propose RouTex, a novel multi-reference 3D texture generation framework based on in-context Diffusion Transformers (DiT). Our key insight is that native 3D Rotary Position Encoding (RoPE) inherently possesses spatial addressing capabilities. We introduce Positional Routing (PR) to reprogram 3D RoPE along spatial and semantic dimensions, establishing explicit reference-to-view correspondence without introducing additional parameters. To execute this routing efficiently, Dual-Granularity Attention (DGA) combines per-view local attention for fine texture fidelity with global joint attention over deduplicated references for cross-view consistency. Extensive experiments demonstrate that RouTex achieves new state-of-the-art performance in multi-reference 3D texture generation.
High-content multi-channel imaging, from Cell Painting assays to remote sensing, is central to modern scientific pipelines. Because these channels are highly heterogeneous and experimental configurations frequently evolve, a practical vision backbone must be fundamentally channel-agnostic. Current channel-adaptive vision transformers rely on global self-attention with learnable channel embeddings, which can be computed only for previously known channels. Attempts to overcome this limitation of new channels either discard the embeddings, losing channel identity, or process each channel independently, thereby eliminating critical cross-channel interactions. To address this, we introduce ChanSFormer, an efficient Vision Transformer that replaces rigid embeddings with disentangled spatial-channel attention and individual channel CLS tokens. This architecture natively preserves both distinct channel identities and cross-channel information flow. Furthermore, the individual channel class token design unlocks more informative feature representations for self-supervised learning and enables representation-based channel sampling while remaining channel-agnostic. We evaluate ChanSFormer on two biology multi-channel datasets, CHAMMI, JUMP-CP and a satellite dataset, So2Sat. Experimental results show that ChanSFormer outperforms state-of-the-art methods by 3.93%-4.27% on JUMP-CP in classification accuracy. It also demonstrates exceptional cross-dataset transferability under a self-supervised setting, outperforming previous methods by up to 2.72% in accuracy in the same classification tasks. Furthermore, the disentangled attention reduces the quadratic complexity of channel-by-spatial sequence length of the global attention baseline, thus improving throughput by 55%-313%. Code is available at https://github.com/cvlab-stonybrook/ChanSFormer.
An ordered sequence of end-effector poses can give a redundant robot arm multiple valid joint configurations at each waypoint. Inverse kinematics (IK) computes these configurations. Choosing a configuration only for the next waypoint can leave the arm poorly placed for later waypoints. This may increase planned motion time or cause a later planning failure. This paper presents Future-Window Inverse Kinematics (FWIK), an online method that selects IK solutions by considering a small window of future waypoints. FWIK generates several IK candidates, connects them in a layered joint-space graph, and ranks candidate paths with a simple motion-time estimate. A learned model removes paths whose transitions are likely to cause self-collision. A standard MoveIt 2 planner then checks the next-waypoint IK goals from the remaining paths in ranked order, and the first goal that yields a valid trajectory is selected for execution. We evaluate FWIK in simulation on two 7-DoF manipulators, the Franka Emika Panda and the KUKA LBR iiwa. Each method-planner setting contains five repetitions of 100 sequences with 50 waypoints. Across Pilz, RRT-Connect, CHOMP, and STOMP, FWIK completes 96.8-100% of sequences and gives the shortest planned trajectory duration in all eight robot-planner settings. Its mean duration is 12.8-13.1% lower than STAMPEDE, 22.5-25.6% lower than IKLink, and 27.5-37.2% lower than retry-based TRAC-IK. The results show that a small future window can improve sequence completion and reduce planned motion time while leaving final trajectory validation to the motion planner.
Large language model (LLM) shopping assistants now recommend products in multi-turn conversations. However, reliable evaluation faces Data scarcity and Difficulty Standardizing Criteria. To address these obstacles, we introduce CARTBench for CART (Complex and Ambiguous Requests with Timelines). For Data scarcity, we curate cases from real commercial shopping needs across 17 retail domains, pairing timestamped dialogue with 1,603 product records. For Difficulty Standardizing Criteria, we develop case-specific rubrics guided by Ideal Response Principles, distilled from excellent replies manually selected from successfully completed orders. These principles also guide Principle-Guided CART Prompting, our structured recommendation strategy. The experimental benchmark contains 1,328 expert-annotated criteria across Core, Excellence, and Safety. To answer the CART question, we evaluate 7 flagship commercial API models, including GPT-6.1-Sol and Claude-Opus-5, and 35B+ parameters open-weight models. In the first full-suite evaluation under the frozen rubrics, 3 of 10 models exceed 90% on Core, while none exceeds 45% on Excellence. Therefore, we identify an understanding-to-execution gap between required behavior and high-quality purchase advice. To examine these limitations, we diagnose shared failures in product evidence and unresolved needs, linked to changing requirements and decisive unknowns. Code and data are available at https://github.com/Da1yuqin/CARTBench.
EEG-based brain-computer interfaces support a broad range of applications, yet designing decoding architectures that perform well across diverse tasks remains challenging. We introduce AutoBCI, an agentic framework in which a Designer Agent and a Forecaster Agent support the discovery and selection of EEG decoding architectures across tasks. The Designer Agent performs Pool-Guided Architecture Discovery (PGAD), generating and refining architectures through training and validation across multiple EEG tasks, such as emotion recognition, motor imagery, and sleep staging. The Forecaster Agent performs Performance Estimation from Early Knowledge (PEEK), using architecture code, the training protocol, and early learning curves to predict full-budget validation performance and select promising candidates for continued training. Across 14 EEG datasets spanning motor imagery, emotion recognition, and sleep staging, we evaluate AutoBCI with six LLMs, including Opus 5.5 and GPT 5.6 Sol, and compare the architectures selected by the search procedure against ten baselines: six conventional EEG models and four foundation models. The architecture discovered by AutoBCI with Claude Opus 5.5 achieves 64.16% average test balanced accuracy (bAcc), compared with 63.87% for REVE, the strongest baseline on this metric. Using ten observed epochs, PEEK reduces mean absolute error in predicting average validation bAcc from 2.20 to 1.36 percentage points, a 38.1% reduction relative to the best-observed-score baseline. Code is available at https://github.com/Jiang-Muyun/AutoBCI
Hanzhang Qin, Sophie H. Yu, Yuewei Zhang, Zhi-Hai Zhang
Firms using large language models under fixed budgets must decide which model to call and whether to retry a failed task. Because a call's charge is revealed only after completion, one long response can exhaust the remaining budget and leave later tasks unserved. Our objective is to maximize the number of tasks solved before service stops. We develop a contextual bandit policy, ADP-HT, which learns each model's task-specific success probability and constructs conservative cost estimates from coarsened response-length labels on a quantile grid with $2^K$ intervals. The policy weighs predicted success against these costs, adjusting the cost penalty to realized spending and the budget available per remaining task. Under calibration, local regularity, and grid-coverage conditions, we bound the excess allowance in the policy's conservative cost estimates. At a fixed sample size, increasing $K$ reduces the discretization term but increases the estimation allowance. At the tightest tested budgets, ADP-HT solves more than twice as many MATH500 and GPQA tasks as the best implemented baseline. It achieves the highest mean cumulative score at ten of fourteen RouterBench budgets. On LiveCodeBench, it outperforms a deployed commercial router in both default and cost-saving modes at the three lowest test budgets.
Rohan Agarwal, Robin Andeer, Alex Armbruster, Bennett Arthur, Michail Christos Doukas, Piero Esposito, Qiming Fang, Aron Filkey 等 33 位作者
Conversational video characters must generate video as fast as it plays, for one user at a time: at 24 fps, a sustained 166.7 ms per four-frame chunk at batch size one, with only a few chunks of buffering. We describe the system behind Runway Characters, which serves a $\approx$57B-parameter autoregressive video diffusion transformer (four denoising passes per chunk) at 704$\times$1088 and 24 fps on four H100 GPUs with a voice agent in the loop. Each chunk takes 240 attention calls, each reading a 5,788-slot KV cache in which only 757 tokens are new, and 29% of the original step was whole-cache work repeated in every call. We move that work to write time (a ring cache, RoPE and FP8 quantization applied once per write, and a delta rotation when positions reset), fuse the tensor-parallel all-reduce into the FP8 GEMM with in-switch reduction, and give the decoder that turns latents into video frames its own GPU. Without the fused kernel, our optimizations cut the diffusion step from 240.9 to 159.3 ms (1.51$\times$, 24.5 fps sustained) with no detected degradation in lip-sync scores; the fused kernel, rolled back once in production and since re-enabled, reaches 147.4 ms (1.63$\times$, 43% of FP8 peak). In seven weeks of production the worker averaged above 23.5 fps in 99.5% of sessions that streamed for at least a second, and our pipeline takes 0.49 s of a 2.1 s median conversational turn. On the same GPUs, the two open-source real-time video systems we benchmarked, at four denoising passes, miss real time or add over a second of latency (one by estimate), at a third to half of our resolution.
AI 速览：作者提出让编码智能体在仿真中重建任务并生成带动作标签的逼真多视角数据，以替代人工遥操作采集机器人示范。
Teleoperation remains the default way to collect robot demonstrations, but every episode costs robot and operator time, and every new task starts from scratch. We ask whether a coding agent can take over the teleoperator's job. In our DataFarm, the agent receives a data request and one calibration episode of the robot in the target scene and rebuilds the task in simulation: it writes the scene, an expert controller, and a success judge. The simulator supplies exact actions, and a video model conditioned on real images renders each rollout as photographic video, generating all camera views together consistently. The output is real-ready data: synchronized camera views labelled with the robot's actions, on which a policy can be trained directly and deployed on the real robot. Independent judges certify the expert and every rendered episode before admission. On five real-robot tasks, policies trained only on farm data match policies trained on as many teleoperated episodes (93 vs. 84 successes in 100 trials), and under an unseen tablecloth and coloured light they keep 87 successes where the teleoperation-trained policies fall to 46. Adapting a task is fully automatic and costs \$21 of API usage on average, paid once per task, plus \$5 for a render contract; every further episode costs \$0.80 on average for simulation, rendering, and review, with no robot or operator. One hour of three-view data for a task costs \$107 on average, these costs included, slightly below a quote of \$117 for teleoperation in China and below \$290 in labour alone in North America.
AI 速览：为共享本地主机上的多个LLM智能体提供Shell MCP服务器，通过资源感知调度错峰执行命令，缓解资源争用并提升利用率与吞吐。
Local LLM agents often work on shared hosts. Unlike multi-tenant cloud clusters, these hosts often have tightly bounded memory, CPU, and I/O resources. Concurrent agents frequently suffer from resource contention, leading to failures such as OOM and reduced resource efficiency. Considering that agents' resource demands are bursty and shell command delay is imperceptible to agents, there is a huge scheduling opportunity to arrange when and in what order commands are actually executed. Thus, we present ToolSched, a Shell MCP Server for shared local hosts that serves multiple agents and schedules their operations. ToolSched embeds a resource model into the MCP interface: agents estimate resource demands along with command requests. These hints, together with real-time host resource conditions, determine when commands are actually executed. Internally, ToolSched applies a scheduling algorithm to stagger commands, reducing contention and improving resource utilization. Prediction correction and checkpoint-based recovery mitigate estimation bias and resource errors. Our evaluation on Terminal Bench shows that, compared with unscheduled execution, ToolSched improves resource utilization by 148.4% and end-to-end command throughput by 29.4%, demonstrating its effectiveness and practicality.
Long-horizon reasoning remains a central challenge for large language models (LLMs) under sparse-reward regimes. We argue that this brittleness arises from two biases induced by complex reasoning spaces: an exploration bias, where models are drawn toward locally plausible but structurally unstable branches, and a compounding bias, where small local deviations accumulate across depth and suppress rare rewards. We introduce Symbolic Closure Analysis (SCA) as a theoretical lens characterizing how branching structures and sparse rewards induce these biases in long-horizon reasoning with local admissibility, and as a design principle for structural priors in less formal reasoning tasks. Motivated by this analysis, we propose SAGE (Structural Admissibility-Guided Exploration), a unified framework that injects structural guidance to alleviate exploration bias and compounding bias in long-horizon reasoning. SAGE combines two complementary structural guidance: algebraic sparsification, which projects locally admissible candidates onto operator-indexed algebraic subspaces to suppress spurious branching and mitigate exploration bias, and hyperbolic structural guidance, which embeds reasoning states into a negatively curved space to provide dense depth-wise signals and mitigate compounding bias. Across 12 benchmarks and 7 model families, SAGE outperforms competitive baselines. In particular, SAGE achieves up to an 8-fold improvement on the Andrews-Curtis problem, an open real-world long-horizon task. Code is available at: https://github.com/Susan571/SAGE-NeurIPS2026.
Lei Zhang, Ju Dong, Kaixin Bai, Minheng Ni, Yu Fu, Zoltan-Csaba Marton, Zhaopeng Chen, Jianwei Zhang
Safe task-level manipulation requires a planner to recognize hazards, select appropriate mitigation actions, and satisfy their execution prerequisites without abandoning the original goal. We introduce ResponsibleRobotBench, a benchmark of 23 manipulation task configurations with separately specified goals and safety conditions. It connects hazard-category predictions, planned actions, execution feedback, and terminal outcomes at the episode level. Across evaluations of five large multimodal models, terminal safety and safe task completion can diverge: defense prompts increase terminal safety for three tested models but affect safe success in different directions. Correct hazard-category predictions also fail to ensure executable mitigation. In a cutting task under the default configuration, GPT-4o mini identifies the human hazard in all 100 episodes but never requests assistance, and every episode ends unsafe. Offering a help action and actually requesting it lead to different comparisons: outcomes differ between action menus even for a model that never requests help, and repeated requests can leave execution failures unresolved. Imitation-learning evaluations and physical robot sequences illustrate the need to coordinate mitigation with goal actions. Together, these results motivate evaluating mitigation selection, action prerequisites, and recovery alongside hazard recognition and terminal outcomes. The appendix provides task definitions, checker conditions, and prompts. Project website: https://responsible-robot-benchmark.github.io/.
Ju Dong, Xinyu Xiao, Yiming Zhou, Tianshuo Yang, Xuerong Liu, Lei Zhang, Jian Chen, Haocheng Zhao 等 15 位作者
Humanoid loco-manipulation requires coordinated body motion and hand interaction over time, and a plausible-looking command sequence can still drop a held object or leave an interaction incomplete. We present action-conditioned latent foresight, a latent world-action model (WAM) that evaluates proposed command sequences before execution. An action generator proposes joint motion-hand command sequences from vision, language, and the robot state. An action-conditioned endpoint predictor estimates the visual features each sequence prefix reaches and measures progress toward a demonstration-derived subgoal, while a joint temporal prior scores temporal consistency given the robot state and preceding commands. The robot executes the first commands of the best-scoring sequence and replans from the next observation. During training, action consistency regularization reconstructs the first demonstrated command from observed and predicted feature changes, encouraging the predictor to retain action-relevant information. On ten tasks in the humanoid loco-manipulation simulation benchmark SIMPLE, our method achieves 89.7% average success, compared with 73.7% for Psi-Zero and 82.0% for DiT4DiT. On seven real-robot tasks, it achieves 85.7% under standard conditions and 65.7% under unseen object configurations, while GR00T N1.7 achieves 70.0% and 48.6%, respectively. Project page: https://ju6276.github.io/Action-Grounded-Foresight/
Low-quality facial videos suffer from missing fine details, while AI-generated videos may additionally contain implausible teeth and lip movements that do not match the speech. Restoring the mouth region is particularly challenging: enhancing existing details can preserve structural errors, whereas regenerating them risks changing facial identity and articulation. We propose an audio-guided facial video restoration framework that reconstructs unreliable mouth details and aligns lip movements with the speech while preserving facial identity and temporal consistency. Built on a video diffusion transformer, our framework mixes input video latents with noise to retain useful source information while allowing corrupted details to be regenerated. Audio guides lip movements, reference images constrain facial appearance, and previously generated frames promote temporal consistency. We train the model in two stages, first learning visual restoration from degraded images and videos, then learning audio-guided lip correction from degraded dubbed videos. The resulting model supports both restoration with the original audio and visual dubbing with replacement audio. On talking-head restoration benchmarks, our method achieves the lowest LPIPS and DISTS and the best lip synchronization scores among the evaluated restoration baselines. Qualitative results on AI-generated videos further show improvements in facial detail and mouth-region appearance, including clearer tooth boundaries and lip contours.
LLM agents increasingly rely on reusable skills to plan tasks, invoke tools, recover from errors, and complete workflows without modifying model parameters. Although these skills are typically developed to enhance benign tasks, they may also make an agent more capable of executing procedures for harmful purposes. We study this risk, which we call skill-induced behavioral misalignment: adding benign skills may cause an agent to execute external actions that violate safety requirements. We first conduct a systematic study to validate the existence of this phenomenon. It compares agents with and without benign skills across five LLM backbones and three benchmarks under matched requests. On average, skill augmentation improves benign task accuracy by 10.3 percentage points (pp), but also increases unsafe action rate (UAR) by 10.3 pp. The regression also occurs with skills curated by experts. A paired trajectory analysis shows that 24 of 50 newly unsafe trajectories fail to complete the harmful objective, revealing violations that task completion alone would miss. We further disclose multiple factors and scenarios that could exacerbate this risk. (1) Evolving skills with benign feedback can amplify unsafe execution; (2) This regression is more pronounced when retrieved skills are matched to task and contain procedural guidance; (3) Distilling benign experience into reusable skills yield higher UAR compared with their source trajectories. (4) A guard that checks proposed tool actions before execution reduces violations while retaining gains on benign tasks. All these findings confirm the significance of skill-induced behavioral misalignment, and the urgent need for new mitigation solutions.
Diffusion Large Language Models (dLLMs) have emerged as a compelling paradigm for language modeling, offering the unique potential for highly efficient inference via multi-token parallel decoding. However, unlocking this potential remains challenging. High-quality generation typically defaults to one-token-per-step static decoding, while existing parallel algorithms rely on heuristic confidence thresholds, introducing a fragile trade-off between decoding efficiency and generation quality. To overcome this, we introduce Free Draft-and-Verification (FreeDave), a novel training-free and model-free fast decoding algorithm. FreeDave capitalizes on the inherent properties of dLLMs by treating future token predictions as ``free drafts'' and rigorously self-verifying them in subsequent forward passes. Theoretically, FreeDave searches for a shorter decoding path that is guaranteed to reproduce the exact sequence generated by static decoding, achieving algorithmically lossless acceleration. With a sufficiently large draft budget and a verification frontier that does not retreat, it can use the fewest possible model forward passes among feasible paths. Extensive evaluations on math reasoning and code generation benchmarks demonstrate that FreeDave can boost the number of generated tokens per model forward (TPF) by up to 5.48$\times$ and tokens per second (TPS) by up to 3.94$\times$, while robustly maintaining generation quality and avoiding the task-specific performance degradation faced by heuristic methods. Code is available at https://github.com/cychomatica/FreeDave.
Long-context inference is bound by the KV cache, which 2-bit codes could shrink about sixfold if their error stays in check. The usual remedy rotates keys before quantizing, assuming rotation always helps. However, we observe that it does not: rotation flattens a spiky piece of a key but can give a flat one a new peak, and both coexist within one key. We therefore propose FlagRot, which stores each 32-channel piece rotated or as is, whichever has the smaller calibrated attention error, and rotates the query instead when reading. We further show that how often to decide follows the quantization axis: per-channel caches such as KIVI need one decision per head, at no write cost, while per-token caches decide per key, with the choice stored in existing sign bits. At 2.5 bits per value, FlagRot lowers the twelve-model KL divergence of KIVI from 18.7 to 12.4 x 10^-3, below RotateKV on 9 of 12 models. Served as a vLLM plugin, it raises RULER at 128k tokens by +2.9 points and the LongBench v1 mean by +0.4 to +1.9 points on the three of four models where 2-bit KIVI loses accuracy. It also lowers the error of 4-bit integer and FP4 caches.
Zhiwei Zhang, Wenxuan Yuan, Xiaoyuan Cheng, Pouria Behnoudfar, Chuanqi Chen, Charlotte Moser, Yuta Tarumi, Jin-Long Wu 等 13 位作者
Koopman representations describe nonlinear dynamics through linear evolution in a transformed space, offering an attractive framework for their linear simplicity and generality. Yet these advantages do not necessarily translate into improved performance in deep latent data assimilation (LDA). We systematically compare seven representative methods and structural baselines across five systems spanning approximately periodic to strongly chaotic dynamics and a broad range of state dimensions. We evaluate forecast and assimilation accuracy, physical state recovery, latent-trajectory behavior, and dynamical consistency, and assess Gaussian error assumptions, objective convexity, and uniqueness of analysis solutions. Our results reveal a trade-off between representational flexibility and optimization tractability. Linear evolution of the latent state can stabilize short-term forecasts but restrict expressiveness in strongly chaotic and high-dimensional systems, without consistently improving assimilation accuracy. Parameterized linear operators offer greater flexibility, yet their joint optimization with the encoder--decoder remains nonconvex. Methods with closed-form analysis updates generally perform better and exhibit greater robustness in our experiments, although matrix inversion remains a computational bottleneck. These findings show that the practical benefits of Koopman structure depend on both sufficient observational information and the choice of assimilation procedure.
Streaming digital humans produce detailed video while continuing speech as new text arrives. High-resolution denoising is costly, and text updates can cause repeated or interrupted speech. We present TaoMate-HD, an autoregressive model for high-resolution audio-video streaming. For each short segment, or chunk, two denoising steps establish coarse video structure at low resolution. A final step refines the video to high resolution, with audio generated throughout the same trajectory. Each stage passes its audiovisual predictions to the same stage of the next chunk as history. Across text updates, the model retains this history to infer what has already been spoken. Continuation training utilizes asymmetric text conditioning, with separate text windows for generation and distillation. Experiments show a 1.6x speedup over direct high-resolution generation with the same model and hardware. Optimized stage-parallel deployment provides a further 1.9x overall speedup over serial progressive generation. Our continuation training reduces character error rate by 11.18 percentage points relative to the baseline. The model also supports hour-long audiovisual generation. Our project page is https://taoliveaigc.github.io/TaoMate-HD.
Jiawei Zhang, Huachuan Qiu, Ruocheng Guo, Zijun Shi
LLM-based simulated interviewees support interviewer training and the evaluation of mental health interviewing systems. Existing studies have primarily focused on improving human-likeness through richer patient profiles and behavioral modeling. However, human-likeness alone does not establish their reliability. Reliable simulation requires preserving profile-grounded clinical facts (informational invariance) while adapting disclosure and interaction behavior to the interview (relational adaptivity). To this end, we introduce PsyInterviewee, which separates a protected truth channel from a dynamic interaction channel and verifies generated clinical claims against the profile. Specifically, asymmetric write permissions allow dialogue context to update disclosure and behavioral states without authorizing new clinical facts; these states guide what profile-supported information is revealed and how it is expressed. In an automated evaluation of interactions between LLM-based interviewers and interviewees, PsyInterviewee achieves higher scores than four representative baselines on most metrics across both axes, as assessed by two LLM judges. Ablations suggest that the two channels contribute to clinical grounding and response modulation, respectively.
Penetration testing is essential for assessing and strengthening system security against real-world threats, yet traditional workflows remain highly manual, expertise-intensive, and difficult to scale. Although recent advances in Large Language Models (LLMs) offer promising opportunities for automation, existing applications rely on simplistic prompting without task decomposition or domain adaptation, resulting in unreliable black-box behavior and limited insight into model capabilities across penetration testing stages. To address this gap, we introduce PentestEval, the first comprehensive benchmark for evaluating LLMs across six decomposed penetration testing stages: Information Collection, Weakness Gathering and Filtering, Attack Decision-Making, Exploit Generation and Revision. PentestEval integrates expert-annotated ground truth with a fully automated evaluation pipeline across 346 tasks covering all stages in 12 realistic vulnerable scenarios. Our stage-level evaluation of 9 widely used LLMs reveals generally weak performance and distinct limitations across the stages of penetration-testing workflow. End-to-end pipelines reach only 31% success rate, and existing LLM-powered systems such as PentestGPT, PentestAgent, and VulnBot exhibit similar limitations, with the evaluated autonomous agents showing particularly low success rates under our benchmark settings. These findings highlight that reliable autonomous penetration testing requires stronger structured reasoning and fine-grained stage-level diagnosis. PentestEval provides the foundational benchmark needed for future research on fine-grained, stage-level evaluation, paving the way toward more reliable LLM-based automation.
Routing-driven timely data collection in Underwater Acoustic Sensor Networks (UASNs) is crucial for marine environmental monitoring, disaster warning, and underwater resource exploration, etc. However, harsh underwater conditions, including high delays, limited bandwidth, and dynamic topologies, make efficient routing decisions challenging in UASNs. In this paper, we propose a smart interrupted routing scheme for UASNs to address dynamic underwater challenges. We first model underwater noise influences from real underwater routing features, e.g., turbulence and storms. We then propose a Software-Defined Networking (SDN)-based Interrupted Software-defined UASNs Reinforcement Learning (ISURL) framework, which ensures adaptive routing through dynamical failure handling (e.g., energy depletion of sensor nodes or link instability) and real-time interrupted recovery. Based on ISURL, we propose the MA-MAPPO algorithm, integrating multi-head attention mask mechanism with MAPPO to filter out infeasible actions and streamline training. Furthermore, to support interrupted data routing in UASNs, we introduce MA-MAPPO\_i, MA-MAPPO with interrupted policy, to enable smart interrupted routing decisions in UASNs. The evaluations demonstrate that our proposed routing scheme achieves exact underwater data routing decisions with faster convergence speed and lower routing delays than existing approaches.
LLM agents are rapidly shifting software development toward intent-driven \emph{vibe coding}, yet their effectiveness drops sharply in high-level synthesis (HLS), where code optimization must jointly consider design structure, directive search, and downstream hardware constraints. Existing HLS optimization approaches, including both heuristic-based methods and recent LLM-assisted pipelines, typically operate at fixed abstraction levels or in weakly coupled stages, limiting their ability to leverage the strong interplay between source code restructuring and directive search outcomes. Our key insight is that HLS optimization can be reframed from a fixed, hardwired transformation pipeline into a skill-based agentic control problem: an LLM agent equipped with modular, executable optimization skills and a feedback-driven orchestrator can autonomously navigate the coupled space of code restructuring and directive search. This motivates the design of HLSClaw, the first HLS-specific, skill-based agentic harness for holistic HLS optimization. On real-world HLS benchmarks, HLSClaw achieves a 39.12x geometric-mean latency speedup, outperforming the compiler baseline ScaleHLS by 1.98x, and surpassing the commercial coding agents Claude Code and Codex by 1.86x and 1.70x, respectively, while maintaining strong timing robustness.
Penghong Zhao, Yihao Chen, Maoheng Yuan, Yunjian Zhang, Abudukelimu Wuerkaixi, Haoran Xu, Man Lv, Yaolong Wang 等 11 位作者
Reinforcement learning for long-horizon LLM agents faces a fundamental dilemma: successful experiences are essential for policy improvement, yet limited policy capabilities make such experiences difficult to discover. Sparse outcome rewards compound this challenge, leaving sampled trajectories with little relative feedback for learning. In this paper, we propose TriGRPO, a tree-based reinforcement learning framework that turns scarce successes into exploration and learning signals through hindsight credit assignment within and across rollout trees. TriGRPO constructs rollout trees through hindsight-guided branching: when an ongoing trajectory reaches a matched decision state, a successful action discovered during rollout initiates a sibling branch that continues autonomously, enabling further exploration and parent-sibling comparisons. Three complementary groups organize policy optimization over these trees: root groups capture within-prompt relative returns, node groups assign credit from parent-sibling outcome differences, and cross groups supplement sparse feedback with historical performance on related tasks. Experimental results demonstrate consistent gains in task performance and rollout efficiency across model scales, highlighting structured hindsight credit assignment as an effective approach to long-horizon agent learning.
Realizing Embodied AI for Science requires robots to translate experimental goals into reliable physical procedures, reducing hazardous and repetitive human work while generating evidence for scientific discovery. Although language-based planning and learned manipulation provide complementary capabilities, their integration requires robots to adapt subsequent operations to actual execution outcomes. For example, an unsuccessful placement changes what must be done next, even when the experimental goal remains unchanged. Existing benchmarks assess planning, manipulation, and laboratory workflows through task completion and reasoning or execution metrics; however, aggregate performance leaves the connection between outcome-dependent choices and physical execution unresolved. To address this evaluation gap, we introduce Labference, a benchmark for evaluating how observed physical outcomes guide subsequent choices and how those choices translate into robotic laboratory actions. Its evaluation protocol is designed to test outcome-dependent behavioral adaptation through controlled evidence variations, while independently measuring operation selection, physical execution, and task completion. Matched Oracle-assisted continuations from shared checkpoints then assess which forms of support improve subsequent choices and task completion. Initial weighing evaluations reveal target-selection errors in the tested VLA policies despite successful object transfers, whereas GPT-driven control achieves correct selection but encounters execution failures and substantial computation costs. Together, these findings highlight the need to connect task-appropriate reasoning with reliable, timely action. By making this connection measurable and diagnosable, Labference supports targeted improvements toward autonomous experimentation for Embodied AI for Science.
Speculative decoding speeds up LLM serving by drafting tokens that the target model verifies in one pass, and its gain grows with the drafts the target accepts. Structured outputs should be an ideal case, because much of a record follows from relations among its fields: an itinerary's days follow from its cities' stays and flights. Serving engines, however, draft such text by proposing it from the prompt or a learned head, and their grammar masks enforce only syntax, so the model must satisfy the relations, which it violates: on NATURAL PLAN's trip problems, Qwen3-8B writes a valid itinerary for 10%, and retries that name the violated relation raise this to 14%. We present FactorSpec, which serves exact relational constraints as speculative drafts. FactorSpec compiles a JSON Schema and its relations into a factor graph with an exact feasibility check, drafts every token the constraint determines as certain, verifies each request at a width chosen for it, and computes the masks in worker processes outside the scheduler while the GPU runs. In SGLang, its drafts yield up to 10.5 tokens per verify step, up to 1.9x those of the fastest n-gram drafter, and it decodes up to 1.43x faster than speculators tuned per load, at a grammar mask's cost; it also leads in vLLM. Because every record is also valid, it serves up to 13.9x the valid records per GPU-second of syntax-only baselines.
In an era defined by the explosive growth of data and rapid technological advancements, Multimodal Large Language Models (MLLMs) stand at the forefront of artificial intelligence (AI) systems. Designed to seamlessly integrate diverse data types-including text, images, videos, audio, and physiological sequences-MLLMs address the complexities of real-world applications far beyond the capabilities of single-modality systems. In this paper, we systematically sort out the applications of MLLM in multimodal tasks such as natural language, vision, and audio. We also provide a comparative analysis of the focus of different MLLMs in the tasks, and provide insights into the shortcomings of current MLLMs, and suggest potential directions for future research. Through these discussions, this paper hopes to provide valuable insights for the further development and application of MLLM.