#01Oct 19, 2026
cs.RO
Where is Physical RSI capability from? An Observational Scaling Law of LLM-as-policy
Zekai Li, Jiayi Yu, Jinlu Zhang and 8 more
In this paper, we systematically investigate the mechanisms, scaling behavior, and generalization of physical Recursive self-improvement. We introduce PhysRSI-eval to quantify physical RSI through a five-axis metric: First-Attempt Performance, Efficiency, Success, Budget and Improvement from failure. Different VLMs are evaluated on 18 manipulation tasks under a fixed interaction budget. The analysis of experimental results motivates the Physical Barrel Hypothesis: different capabilities may constrain different parts of improvement. The five-axis metrics also distinguish systems with close public benchmark averages and show that high scores need not come with early or efficient gains. We study observational scaling from two perspectives: predicting the relative performance of VLM agents from their public capabilities, and forecasting later progress from early physical feedback. Together, these analyses ask how quickly and how far an agent can improve. Moreover, we conduct workspace interventions and then analyze what improvement leaves behind. Removing history reduces late-round success, frozen workspaces remain useful on unseen layouts, and three cross-task cases achieve success earlier. The results suggest that scene-specific adaptation and reusable experience can coexist.
#02Oct 19, 2026
cs.CV
Can It Actually Work? Synthesizing Functional CAD Mechanisms
Wenqian Zhang, Yangyi Huang, Zigeng Xu and 2 more
While LLM agents can generate visually plausible CAD assemblies, these designs often fail to fulfill their intended physical functions. We study functional CAD design, a setting where agents interpret underspecified requests to construct and iteratively revise mechanisms that satisfy functional requirements in physical simulation. Verifying these requirements is nontrivial: different candidate mechanisms may realize the same function through entirely different components, controls, and operating sequences. To bridge this gap, we propose FunCAD-Harness, a framework that translates abstract requirements into candidate-specific physical tests and uses simulation feedback to guide design revisions. To systematically evaluate this capability, we introduce FunCADBench, a benchmark spanning 46 mechanism families with interdependent and multi-step operations. Experiments across frontier LLM agents show that FunCAD-Harness exposes subtle mechanical failures invisible to visual inspection, substantially improving functional success rates over standard simulation-augmented baselines. Ultimately, our work marks a step toward CAD agents capable of translating human intent into working mechanisms.
#03Oct 19, 2026
cs.AI
Same Simulation, Different Evidence: Where Retrieved Tool Output Costs LLM Agents Accuracy
Ke Zhang, Sahchit Chundur, Jerry Peprah Owusu and 2 more
LLM agents that compute the right answer can still answer wrongly. Agent harnesses increasingly keep long tool outputs out of the context window, saving them to files that the agent searches or reads on demand. We ask where this costs accuracy with a real scientific tool: PHREEQC, a widely used geochemistry simulator from the USGS, whose reports reach 846.8k characters and whose determinism lets every printed value be traced. Agents use it to answer 158 execution-verified multiple-choice items. When agents read the simulator's report through a section index rather than in context, GPT-5.1 and GPT-5.2 lose 11.3 and 7.9 points of accuracy, while Claude Opus 4.6 and Claude Sonnet 4.6 show no detectable cost. Re-running all 11,007 logged simulator calls shows that both GPT models compute the keyed value about as often as before; what changes is the evidence they answer from. Under retrieval they often skip the report, answer from their own higher-precision output files, and misround to a neighboring option, a failure we call evidence substitution. With the simulation held fixed, retrieval still costs GPT-5.2 9.4 points. For GPT-5.4, more frequent correct simulations mask an evidence-use loss of about ten points, leaving a net change indistinguishable from zero. Naming the authoritative output raises accuracy by 6.3-7.9 points over plain retrieval for the three GPT models; for GPT-5.1 and GPT-5.2, inlining outputs that fit works as well, while asking for correct rounding does not. Context-management choices in agent harnesses should be judged by where accuracy is lost, not only by final accuracy.
#04Oct 19, 2026
cs.CV
VideoMaMa++: Temporally Consistent Video Matting via Preserve-and-Refine Token
Sangbeom Lim, Seoung Wug Oh, Heeji Yoon and 2 more
Diffusion-based video matting has recently emerged as a leading approach, leveraging video diffusion priors from internet-scale pretraining to achieve strong zero-shot generalization from limited synthetic data. However, the heavy compute of these diffusion backbones limits inference to a short chunk of just 6 to 24 frames, so even ordinary videos span multiple chunks. Because each chunk is matted independently from coarse mask guidance, existing methods fail to maintain consistency across chunk boundaries, producing flickering edges and discontinuous transparency. We address this with a mask-guided diffusion-matting framework built on two ideas. The RGB and mask conditioning are represented as separate token streams that interact through full spatio-temporal self-attention, letting the model adaptively combine fine RGB texture with coarse mask layout rather than committing to a fixed channel-wise fusion. On top of this, we propose a novel pair of learnable embeddings, the preserve token and the refine token, that act as a per-frame conditioning interface and enable a sliding-window inference scheme in which already-generated mattes are propagated across chunk boundaries for local temporal consistency, complemented by a globally shared reference anchor at every chunk to prevent error accumulation in long sequences without any additional training. Across standard video matting benchmarks and in-the-wild long footage, our framework outperforms prior diffusion-based methods in both fine-detail accuracy and temporal consistency, and uniquely maintains consistency across chunk boundaries on long videos.
#05Oct 19, 2026
cs.CL
OpenVLHarness: Real World Visual Agentic Tool-Use with Multimodal Memory
Yu Zhou, Cheng-Fu Yang, Rui Sun and 8 more
Multimodal large language models (MLLMs) have made rapid progress on visual reasoning, yet many tasks require evidence that is difficult to obtain through internal reasoning alone: our controlled study on 22 benchmarks shows that simply increasing the inference budget of the base model yields only modest gains. Such tasks demand grounded perception, external knowledge, and numerical measurements obtained through heterogeneous tools, whose outputs must be made interpretable to the model and retained across reasoning steps. We introduce OpenVLHarness, a multimodal reasoning harness organized around a capability layer and persistent multimodal memory. The capability layer combines task-level tool interfaces with evidence rendering, providing consistent access to heterogeneous backends and converting their outputs into model-readable observations. Persistent multimodal memory retains the corresponding images, structured data, and web references as addressable artifacts throughout each query. A coding interface executes generated programs over stored data, enabling computations that combine results from multiple tools. Together, these mechanisms support an iterative process of acquiring, inspecting, and composing multimodal evidence. Evaluated on 22 benchmarks with eight open and proprietary backbones, OpenVLHarness improves average scores by 9.7-12.6 points over matched no-tool baselines, including +11.2 with Qwen3-VL-8B, +9.7 with Kimi K3, and +9.9 with GPT-6 Astra. With GPT-6 Luna at medium reasoning effort, OpenVLHarness outperforms the base model, Codex, and Visual Sketchpad at maximum effort by 7.7-8.5 points while incurring at most 37% of their model-token cost.