Search arXiv⌕ Search

arXiv subjects

Fan Yang

Publications and source records attributed to Fan Yang.

At least 37 records · Page 2Linked to original sources
arXiv2026

Beyond What Is Visible: Context-Aware Sentiment Across Online Communities Among Crisis Events

Social media offers timely insight into disaster experiences, yet online communities may not represent those affected on the ground. Uneven participation can bias crisis information systems toward data-rich locations and obscure communities with fewer digital traces. We address this with an adaptive cross-community learning framework that combines text and aggregate mobility signals with socioeconomic and hazard-informed community representations, supplementing sparse local observations with evidence from structurally related cities. We evaluate the framework on the January 2025 Southern California wildfires and 2024 Hurricane Helene in Florida. The adaptive model achieves the highest F1-score in both cases (0.5923 and 0.7615), while augmentation has its strongest effect where original observations are limited. The framework also reveals finer cross-city community-level sentiment variation and stronger sentiment-mobility correspondence than text-only models, with relationships differing across disaster contexts. These findings support crisis systems that better account for context-dependent evidence and improve interpretation of underrepresented community signals.

cs.CY↗
arXiv2026

SkillScaffold: Scaffolding Agentic Reinforcement Learning with Policy-Adaptive Skill Credit Assignment and Selective Injection

Agentic reinforcement learning (RL) enables large language models (LLMs) to learn from multi-turn interactions with verifiable outcomes. Textual skills can improve exploration by distilling prior interactions into reusable guidance. However, existing skill-based RL approaches often either assume skill injection is beneficial or rely on indirect feedback or rollout-intensive evaluation to assess its utility. Our pilot studies show that rollout-based utility estimation for skill credit assignment is costly and unstable under limited sampling, while skill benefits vary across candidates and depend on the current policy's competence on each query. To address these challenges, we introduce SkillScaffold, a policy-adaptive training scaffold that combines low-cost skill-utility estimation with selective skill injection. By reusing outcome-verified training trajectories, SkillScaffold generates candidate skills and estimates their utility through outcome-aware likelihood comparisons, without candidate-specific environment rollouts. The resulting utility guides whether and which skill to inject, retaining skill-free execution when no candidate is estimated to help. Joint training enables the Generator to produce skills tailored to the evolving Executor. Scaffolded RL encourages skill internalization for skill-free inference. Experiments on ALFWorld, WebShop, ScienceWorld, and $τ^2$-Bench demonstrate gains over strong agentic-RL and skill-based RL baselines across both short- and long-horizon tasks. Ablations and further analyses support the importance of utility-based selection and abstention. Our code is available at https://github.com/Vaidurya00/SkillScaffold.

cs.LG↗
arXiv2026

Who Improves the Robot? Toward Physical Recursive Self-Improvement

Robots are increasingly autonomous in task execution, yet their improvement remains largely human-driven. This study investigates Physical Recursive Self-Improvement (Physical RSI), in which physical experience guides lasting updates to both a robot's capabilities and its improvement procedure, making subsequent improvement more effective across tasks. Physical RSI faces challenges from noisy observations, continuous dynamics, costly trials, and irreversible consequences. We present an L0-L5 framework that differentiates capability-level from meta-level updates and within-task from across-task gains, reserving L5 for independently verified recursive gain. We study existing methods around harness adaptation and model training, alongside safety and control mechanisms that govern execution and successive updates. Our analysis reveals that, despite substantial progress in self-improvement, the surveyed literature does not yet demonstrate Physical RSI as defined by our framework. We identify research priorities in reliable feedback, causal attribution, safety preservation, and benchmarking, outlining a path toward robots that autonomously improve how they learn while ensuring safety. The GitHub repository for Physical RSI is available at https://github.com/SAIL-Research-Lab/Physical-RSI.

cs.AI↗
arXiv2026

RCR-Router: Efficient Role-Aware Context Routing for Multi-Agent LLM Systems with Structured Memory

Large language model (LLM) agents are increasingly organized into multi-agent systems for complex reasoning and decision-making tasks. However, scaling such systems can substantially increase inference cost, especially when multiple agents repeatedly process overlapping or role-irrelevant context. Many existing multi-agent frameworks focus on orchestration and role specialization, but still rely on full-context sharing or static context routing, leading to redundant token consumption, increased runtime, and inefficient memory utilization. We present \emph{Role-Conditioned Context Routing (RCR)}, an execution-time layer that allocates role-relevant shared-memory items to agents under token budgets. Given a structured shared memory and a set of role-specialized agents, RCR selectively routes task-relevant memory items to each agent according to its role, optional task-stage metadata, and token budget. RCR integrates role-conditioned relevance scoring, token-budgeted memory selection, and iterative memory updates to reduce redundant context exposure during multi-agent execution. Experiments on multi-hop QA, embodied reasoning (ALFWorld), and web-based decision-making (WebShop) show that RCR reduces token usage by 35\% on average and runtime by 38\% on average relative to Full-Context, while maintaining or improving task performance.

cs.CL↗
arXiv2026

Tactile Sim2Real without Tactile Simulation via Bottlenecked Latent Reconstruction

Robot sensor designs, particularly tactile sensors, are highly diverse and evolve rapidly. Modeling each sensor in simulation demands substantial domain expertise and computational approximations can degrade the fidelity of the simulated signals. We propose Sim2Real via Bottlenecked Latent Reconstruction (SBLR), a framework that avoids sensor-specific simulation entirely by (1) training policies on a simulator-native oracle sensor that is easy to construct without modeling any particular sensor (e.g. we use a point-cloud and finger-tip forces as a tactile oracle), and (2) aligning real sensor latent embeddings to those of the oracle sensor at inference time. Policy training proceeds in two-stage: the policy first learns from the oracle sensor latents, then a bottlenecked latent reconstruction adapts it to the information loss expected when using the real sensor instead of the oracle. The alignment between oracle and real sensor is learned from unpaired random-play data collected in both simulation and the real world, using rectified-flow-based transformation networks trained on nearest-neighbor pseudo-pairs. Simulation experiments on three contact-rich tasks show that SBLR matches or approaches the performance of an oracle with direct access to tactile simulation. Hardware experiments on Peg Insertion and Gear Meshing with GelSight Mini and DIGIT sensors demonstrate 85-97.5% zero-shot success without requiring any sensor-specific modeling or calibration, outperforming a physics-based tactile simulation baseline by 7.5-15%.

cs.RO↗
arXiv2026

Before Attention Sinks Form: Hidden-State Geometry Reveals and Modulates Their Formation

Attention sinks are usually studied after they have formed. We ask whether a measurable hidden-state change appears beforehand and whether changing that geometry alters what happens later. To remove network depth as a shortcut, we compare prompts in the same model and at the same layer. When sink timing varies across prompts, simple spectral summaries of the hidden states add information beyond current attention, with the clearest prompt-specific signal in a short window about two to three layers before sink formation. The predictive strength varies across cohorts, so we treat the spectrum as a readout of a pre-sink state change rather than a universal warning statistic. The signal is broader than one measured extreme activation. We then modify only the singular values of the post-layer hidden states after current attention has already been computed. These interventions change later sink behavior in Qwen2.5, GLM-Z1, and GPT-OSS relative to norm-matched random spectral perturbations. In Qwen2.5, the immediate suppressive effect is accompanied by a reduction in next-layer query--key preference for position~0, and the learned spectral directions remain stable when estimated from substantially less training data. Together, the results show that attention sinks are preceded by a short spectral reorganization of hidden states, and directly perturbing this geometry can causally alter subsequent sink formation.

cs.LG↗
arXiv2026

Genesis: Visual Augmentation Primitives for Multimodal Chain-of-Thought Reasoning

Recent multimodal large language models achieve promising results on visual reasoning tasks, approaching human-level performance on existing STEM benchmarks. However, these capabilities degrade under complex real-world scenes. We identify that chain-of-thought reasoning remains largely confined to the linguistic space, rarely anchoring each step in visual evidence. The main cause is the absence of such data during training. In this work, we introduce a training data augmentation pipeline, \textbf{Visual Augmentation Primitives} (VAP), which leverages state-of-the-art vision generative models to synthesize diverse photorealistic training samples and corresponding reasoning text. VAP introduces some basic operators and applies them to a seed image and gets a variant. It also constructs coherent multimodal reasoning chains throughout the generation process. Such training data force models to perceive visual content and perform logical reasoning according to them at every step. We further introduce multimodal chain-of-thought rubric-based reinforcement learning to enhance this. It rewards not only text reasoning but also visual reasoning by scoring intermediate reasoning steps for logical validity, completeness, and consistency. Moreover, we construct the UNREAL benchmark to systematically evaluate visual reasoning on hyperrealistic scenes across controllable difficulty levels. Experiments show that a 4B model built on this framework surpasses 8B baselines, with performance improving consistently as the model scales to 32B, validating the scaling law for visual reasoning.

cs.CV↗
arXiv2026

HarnessWAM: Bridging Prediction and Deliberation in World Action Models

World Action Models~(WAMs) jointly learn environmental dynamics and robot actions, introducing priors over physical evolution into embodied control. However, finite-horizon prediction and action generation are insufficient for complex embodied tasks that require global planning, cross-stage state maintenance, execution verification, and failure recovery. We refer to this mismatch as the prediction-deliberation gap. To address this gap, we propose HarnessWAM, an agentic framework for WAMs. HarnessWAM employs a vision-language-model-based Task Manager to maintain an evidence-grounded scene belief and organize the global instruction into a structured task graph. Capability-conditioned executable-space projection then compiles the graph into WAM-supported skill sequences that satisfy task dependencies and embodiment-state constraints. During execution, HarnessWAM couples a lightweight progress estimator with event-triggered Task Manager deliberation. The estimator supplies frequent execution cues, while the Task Manager uses observations, task state, and interaction history to verify outcomes and decide whether to advance, acquire evidence, revise the plan, or recover. Local recovery preserves acquired scene knowledge so that execution can resume after failure. HarnessWAM achieves state-of-the-art average performance on RoboMemArena, with full-task and subtask success rates of 64.4% and 76.8%, respectively, and reaches 23.7% success on RoboCerebra Ideal. On a physical UR3 arm, full-task success reaches 95% for shape-conditioned tabletop tidying and 90% for memory-dependent box exploration. These results demonstrate the effectiveness of model-external agentic orchestration in bridging the prediction-deliberation gap for complex embodied tasks.

cs.RO↗
arXiv2026

VaME: Exploring Variational Latent Reasoning for Multimodal Embeddings

Universal multimodal retrieval requires compact embeddings that preserve task-relevant semantic information across diverse modalities. Prior works have incorporated latent reasoning into multimodal embedding learning to refine this information before embedding extraction. However, most existing approaches remain confined to deterministic latent paths, without exploring alternative trajectories to discover better embeddings. Thus, we propose VaME (Variational Multimodal Embeddings), a framework that models latent reasoning as a learnable distribution over trajectories. Specifically, we first introduce Variational Latent Reasoning (VLR) to enable autoregressive exploration in latent space, guided by answer reconstruction through a lightweight decoder. Meanwhile, we augment the original embedding-token readout with a latent-fused embedding to facilitate exploration during subsequent reinforcement learning. Finally, we optimize latent reasoning over stochastic variational trajectories through reinforcement learning, using Semantic Decoding Reward (SDR) to favor semantically meaningful trajectories with interpretable decoded outcomes. On the 78-task MMEB-V2 benchmark, spanning image, video, and visual-document retrieval, VaME outperforms most explicit CoT-based models and all latent-reasoning baselines. VaME also demonstrates robust performance on reasoning-intensive benchmarks such as MRMR, with substantial gains after reinforcement learning. Importantly, VaME achieves these gains with at least a 4.25x inference speedup over the deterministic latent autoregressive baselines. The code will be made publicly available.

cs.CV↗
arXiv2026

PixVL: Self-Supervised Training of Pixel-Level MLLMs via a Unified Mask--Text Consistency Cycle

Recent studies develop pixel-level multimodal large language models (MLLMs) that support both Region Segmentation and Region Understanding, extending multimodal interaction from whole images to specific objects and regions. However, these methods face two fundamental challenges. First, the scarcity of high-quality mask--text pairs leaves abundant mask annotations without corresponding language supervision. Second, discrepancies in supervision formats and learning-signal densities induce optimization interference between Region Segmentation and Region Understanding. To address these challenges, we propose PixVL, a self-supervised post-training framework that introduces a unified Mask--Text Consistency Cycle, enabling pixel-level MLLMs to generate and self-verify regional descriptions and learn from unlabeled data. We found that direct cycle based solely on geometric reconstruction is unreliable because re-segmentation IoU does not faithfully reflect the semantic quality and referring sufficiency. PixVL therefore introduces confuser-aware semantic verification, which uses the model's confidence when it correctly chooses the target among highly similar candidate regions, and assigns zero reward to an incorrect choice. Meanwhile, PixVL performs cross-view verification using temporally separated video frames or geometrically transformed image views, preventing cyclic learning from collapsing to positional and shape shortcuts. Finally, a quality-coupled bidirectional learning strategy uses the highest-reward description to guide Text-to-Mask learning. This strategy transforms Region Understanding and Region Segmentation from competing tasks into mutual generators and verifiers. Experiments demonstrate that PixVL improves both region understanding task and segmentation task.

cs.CV↗
arXiv2026

From killed BBM with drift to BBM: the extremal process

In this paper, we study the asymptotic behavior of the extremal process of a standard one-dimensional branching Brownian motion (BBM) with drift $-ρ>-\sqrt2$ and absorbing barrier at level $-x$. We prove that the two-dimensional point process, with first component being the extremal process of the BBM and the second component being the running minimum of the BBM with drift, converges weakly to a decorated Poisson point process (DPPP) on $\mathbb{R} \times [0, \infty)$. This framework allows us to supply a common limit coupling of the extremal processes for different barrier levels and to explicitly derive the limit, as $t\to\infty$, of the extremal process of the killed BBM killed at level $-x$, demonstrating that the double limit, when $t\to\infty$ first and then $x\to\infty$, of the extremal process of the BBM killed at level $-x$ coincides with the limit of the extremal process of (un-killed) BBM.

math.PR↗
arXiv2026

APEX: Active Protection at Execution Boundaries for LLM Agents

Indirect prompt injection (IPI) hides adversarial instructions in content that large language model (LLM) agents read at runtime. As agents compose heterogeneous capability units, including Tools, MCP servers, and Skills, the carriers of injection multiply, and defenses built to recognize attack patterns fall behind them. We instead shift defense from covering attack patterns to one stable point: whatever the carrier and however the injection propagates, harm materializes only at the \emph{execution boundary}, where the agent turns internal state into an external action or released output. Safety there turns on two conditions, both settled by the trusted task rather than by the run: whether the proposed effect is authorized, and whether the runtime information reaching it is endorsed by that task. We present APEX, an active defense that enforces both at this boundary from a single authorization contract compiled before untrusted execution: \emph{evidence-gated prevention} admits an effect only when the contract justifies it, while \emph{deception-based exposure} makes unendorsed use reveal itself before the effect commits. Protection therefore follows from what the task permits rather than from how an attack is built, and applies uniformly across capability units without attack-specific policies or taint tracking. Against 13 baselines, APEX attains 0\% attack success on five of six benchmarks and 0.56\% on the sixth, holds 0\% under adaptive attacks on all three capability-unit types, and remains effective across defender backbones. Code is available at https://github.com/ZhengXR930/APEX_official/tree/official.

cs.CR↗
arXiv2026

Effective-Geometry Rescaling and Universal Critical Behavior in the Anisotropic Three-State Potts Model on the Square Lattice

We study the two-dimensional anisotropic three-state Potts ferromagnet on the square lattice using Wolff single-cluster Monte Carlo simulations and finite-size scaling. For coupling ratios $λ=J_y/J_x=0.5$, $0.75$, and $1$, finite-size scaling of the correlation ratio yields critical behavior consistent with the two-dimensional three-state Potts universality class. The scaling of the leading Fisher zeros gives a correlation-length exponent consistent with $ν=5/6$, while their cumulative density is consistent with the expected specific-heat exponent $α=1/3$. We further characterize the anisotropy at criticality using directional correlation ratios $R_x$ and $R_y$ together with directional FK wrapping probabilities. For $λ=0.5$ on a physically square lattice, $R_x$ and $R_y$ approach distinct critical values while yielding a common correlation-length exponent. From the wrapping probabilities, we independently determine an effective aspect ratio $ρ_e^\square=0.6413(5)$, in close agreement with the theoretical value $ρ_e^{\square,\mathrm{th}}\simeq0.64150030$ obtained from the isoradial representation. Using the theoretical value to set the physical aspect ratio restores directional equivalence, $R_x\simeq R_y$, and brings the overall correlation ratio toward the isotropic-square reference. The results show that spatial anisotropy changes the effective critical geometry without altering the bulk three-state Potts universality class.

cond-mat.stat-mech↗
arXiv2026

DexPIE: Stable Dexterous Policy Improvement from Real-World Experience

Dexterous manipulation presents substantial challenges for imitation learning due to its high-dimensional action space and complex contact-rich dynamics. Policies trained purely from demonstrations often suffer from compounding errors during deployment and require large amounts of expert data to achieve reliable performance. To move beyond the limitations of demonstration data, in this work, we propose DexPIE, a post-training framework for dexterous policy improvement from experience collected through real-world deployment. First, DexPIE enables effective exploration coverage through a dexterous-hand-adapted intervention system and multi-stage DAgger-style data collection across initial and intermediate task stages. Meanwhile, we enhance consistency between training and inference to reduce the distribution shift between rollouts and demonstration data, better aligning rollout behavior with demonstrations, allowing the critic to learn a value function induced by a more consistent underlying policy. Together, these components provide reliable supervision for policy evaluation. Finally, DexPIE improves the policy through conditioning on a continuous optimality indicator, allowing the policy to leverage the quality of data in a more fine-grained manner. Across three challenging real-world dexterous manipulation tasks, DexPIE achieves a 37.3% improvement in success rate over the demonstration-based reference policy, outperforming all baseline methods and demonstrating stronger robustness. The source code and dataset will be made publicly available.

cs.RO↗
arXiv2026

What Did the Agent Actually Do? Evidence-Grounded Oversight for Long-Horizon Agents

As agents take on long-horizon tasks, users shift from making individual decisions to overseeing autonomous execution. Yet the volume of agent activity and the fragmentation of supporting evidence make it difficult to determine which decisions warrant user verification. We study monitors that identify consequential decisions and locate evidence to help users assess their implications. We introduce AgentMonBench, a software-engineering benchmark comprising three subsets that cover two complementary dimensions: alignment between requirements and behavior, and awareness of consequential autonomous decisions for verification. To support these judgments, we propose the Evidence-Grounded Behavior Graph (EBG), a training-free method that groups source-linked evidence into behaviors and organizes their relationships into a graph. EBG presents task-oriented views of this graph to help monitors interpret behavior in context. Experiments across eight models show that EBG improves decision identification and evidence localization in most settings compared with direct access to the original context. Further experiments show that EBG's evidence-localization gains persist across input scales and hyperparameter settings, while real-world applications illustrate its practical value for human oversight.

cs.AI↗
arXiv2026

The Functional Structure of Post-Compression Recovery in Low-Rank LLMs

Different low-rank compression methods can produce compressed LLMs that respond differently to the same post-compression recovery procedure, and relative advantages observed between methods at the compression endpoint may shrink, grow, or even reverse after recovery. We ask whether this recovery heterogeneity reflects functional structure beyond scalar loss evolution, and how that structure evolves throughout recovery. Our results establish that this heterogeneity reflects a reproducible compression-induced functional structure, which we formalize as recovery pressure. To characterize this structure consistently throughout recovery, we develop a standardized functional characterization within each backbone that is applicable across heterogeneous low-rank methods. The primary backward characterization reveals reproducible module-wise structure across independent probes, while a complementary forward-only characterization recovers related structure without loss or backpropagation. We further find that recovery pressure measured at the endpoint is associated with subsequent recovery response; during recovery, its module-wise structure is reorganized non-uniformly, and localized updates induce distributed responses beyond directly updated modules. Further evidence indicates that tracking this evolving structure provides a complementary functional view of recovery progress alongside scalar loss.

cs.AI↗
arXiv2026

TRACER: Texture-Robust Affordance Chain-of-Thought for Deformable-Object Region Grounding

The central challenge in robotic manipulation of deformable objects lies in aligning high-level semantic instructions with physical interaction points under complex appearance and texture variations. Existing vision-based affordance prediction methods often suffer from boundary overflow and fragmented functional regions. To address these issues, we propose TRACER, a Texture-Robust Affordance Chain-of-Thought for Deformable-Object Region Grounding framework that maps hierarchical semantic reasoning to appearance-robust and physically consistent interaction regions. Specifically, a Tree-structured Affordance Chain-of-Thought (TA-CoT) decomposes high-level task intentions into hierarchical affordance-semantic instructions. A Spatially-Constrained Boundary Refinement (SCBR) mechanism suppresses prediction spillover and guides responses toward valid object regions. Furthermore, an Interactive Convergent Refinement Flow (ICRF) refines dispersed affordance responses into coherent regions, improving spatial continuity and physical plausibility. Experiments on the Fine-AGDDO15 dataset and a real-world robotic platform demonstrate that TRACER improves affordance grounding precision across diverse textures and patterns. It also enhances long-horizon manipulation success, bridging high-level semantic reasoning and low-level physical execution. The source code and dataset will be made publicly available at https://github.com/Dikay1/TRACER.

cs.RO↗
arXiv2026

Return-to-Home Feasible Micro-Aerial Vehicle Exploration for 3D Gaussian Splatting Reconstruction

Micro aerial vehicles (MAVs) enable rapid indoor 3D reconstruction for inspection and time-critical situational awareness, but must operate under strict flight-time budgets and safety constraints that require explicit return-to-home (RTH) feasibility with a conservative margin. We present a topology-first active reconstruction framework that decouples high-fidelity rendering from navigation. A dense 3D Gaussian Splatting (3DGS) map is optimized as the reconstruction target, while a lightweight Truncated Signed Distance Field (TSDF)/occupancy scaffold supports conservative collision checking and online construction of a sparse 3D Voronoi skeleton roadmap. Since directly extracting roadmaps from TSDF geometry can be unstable under noisy and incomplete online fusion, we validate nodes and edges using carved free-space consistency and dense visibility checks, which suppress behind-wall phantom structure and stabilize planning. Viewpoints are selected on the roadmap using a flight-time-budget-aware objective and a lightweight receding-horizon lookahead to improve non-myopic exploration behavior. We evaluate on photorealistic indoor simulation benchmarks (ReplicaCAD, Gibson, and HM3D), reporting reconstruction coverage/error, RTH success, and computational cost. Results demonstrate improved or competitive performance relative to recent GS-based baselines under identical flight-time budgets.

cs.RO↗