arXiv subjects

Yanzhe Hu

Publications and source records attributed to Yanzhe Hu.

4 recordsLinked to original sources

arXiv2026

TacForcing: Streaming Action Generation with Execution-Time Tactile Feedback

Contact-rich manipulation requires adapting to contact states that can evolve substantially within an action horizon. However, chunk-based vision-language-action models predict complete action chunks from observations collected before execution, leaving tactile conditioning stale during execution. Existing tactile-reactive approaches typically rely on separate high-frequency controllers, which increase both architectural and training complexity. In this paper, we introduce TacForcing, a streaming action-generation framework that effectively incorporates execution-time tactile feedback. Instead of employing a separate reactive controller, TacForcing replaces the standard action expert with a streaming action expert to generate actions conditioned on the evolving tactile observations acquired during execution. TacForcing also introduces Execution-Aware Tactile Attention (EATA), which restricts tactile conditioning to actions nearing execution, thereby reducing the temporal mismatch between tactile acquisition and action execution. Across six simulated UniVTAC tasks and three real-world contact-rich manipulation tasks, TacForcing achieves average success rates of 65% and 69%, respectively, outperforming strong baselines in both settings.

cs.RO↗

arXiv2026

World-Language-Action Model for Unified World Modeling, Language Reasoning, and Action Synthesis

We propose world-language-action (WLA) models as a new class of embodied foundation models. WLA takes textual instructions, images, and robot states as inputs to jointly predict textual subtasks, subgoal images, and robot actions, conjoining the \emph{world modeling interface} to learn from extensive egocentric videos as in the world-action model (WAM) and the \emph{language reasoning} capacities to solve complex long-horizon tasks as in vision-language-action (VLA) models. At the core of WLA lies an \emph{autoregressive (AR)} Transformer backbone, instead of a bidirectional diffusion Transformer as in WAMs, to predict the \emph{next state}, comprising the \emph{semantic-level} textual intention and complementary \emph{fine-grained} physical dynamics. The physical dynamics are supervised by the world modeling objective based on a dedicated World Expert, and are leveraged to ease the characterization of the state-action correlation for the Action Expert. WLA leverages meta-queries to make the world prediction \emph{implicitly} impact the action generation so that the former can be disabled during inference. The world prediction can also be activated to enable test-time scaling for improved robot control. Our WLA-0 prototype, with 2B active parameters, achieves 40 ms per inference on an NVIDIA RTX 5090. Evaluations across simulated and real-world environments demonstrate that WLA-0 achieves state-of-the-art multi-task and long-horizon learning abilities, e.g., 92.94\% success rate on RoboTwin2.0 Clean and 56.5\% success rate on RMBench. WLA-0 also holds the promise to learn novel tasks directly from \emph{cross-embodiment robot videos} without action annotations.

cs.RO↗

arXiv2026

LightningRL: Breaking the Accuracy-Parallelism Trade-off of Block-wise dLLMs via Reinforcement Learning

Diffusion Large Language Models (dLLMs) have emerged as a promising paradigm for parallel token generation, with block-wise variants garnering significant research interest. Despite their potential, existing dLLMs typically suffer from a rigid accuracy-parallelism trade-off: increasing the number of tokens per forward (TPF) via aggressive parallel decoding often leads to performance degradation and increased generation instability. We identify that this limitation stems from the model's inability to navigate high-parallelism regimes where approximation errors and local corruptions accumulate, ultimately undermining the reliability of parallel generation. To address this, we propose LightningRL, a post-training framework designed to directly optimize the speed-quality Pareto frontier of pre-trained dLLMs. Instead of forcing uniform parallelization, our approach leverages reinforcement learning to identify and reinforce high-parallelism trajectories that maintain generation accuracy. Built upon the Group Relative Policy Optimization (GRPO) framework, LightningRL introduces several enhancements tailored for dLLMs: (1) stabilized training via per-reward decoupled normalization; (2) token-level negative log-likelihood (NLL) regularization on correct trajectories to anchor model performance; and (3) a dynamic sampling strategy with TPF-aware filtering to enhance training efficiency. Experimental results across mathematical and coding benchmarks demonstrate that LightningRL consistently advances the Pareto frontier, achieving competitive task accuracy while significantly increasing parallelism, reaching an average TPF of 7.32 (with a peak of 11.10 on the MBPP dataset). Our code is available at https://github.com/SJTU-DENG-Lab/LightningRL.

cs.LG↗

arXiv2025

Unveiling Large Language Model Supply Chain: Structure, Domain, and Vulnerabilities

Large Language Models (LLMs) have revolutionized artificial intelligence (AI), driving breakthroughs in natural language understanding, text generation, and autonomous systems. However, the rapid growth of LLMs presents significant challenges in the security and reliability of the Large Language Model Supply Chain (LLMSC), a complex network of open-source components, libraries, and tools essential for LLM development and deployment. Despite its critical importance, the LLMSC remains underexplored, particularly regarding its structural characteristics, domain composition, and security vulnerabilities. To address this gap, we conduct the first empirical study of the LLMSC, analyzing a curated dataset of open-source packages from PyPI and NPM across 14 functional domains. We construct a directed dependency graph comprising 13,486 nodes, 28,704 edges, and 180 unique vulnerabilities to investigate the structural characteristics of the LLMSC and analyze how security risks propagate through its dependency network. Our findings reveal that the LLMSC exhibits a locally dense, globally sparse topology, with 72.38% of dependency trees containing fewer than 5 nodes, while a few large trees dominate the ecosystem, accounting for 77.66% of all nodes. The graph is characterized by high-degree hubs, with the top 5 most connected nodes averaging 1,207 dependents each. Security analysis shows that critical vulnerabilities propagate to an average of 142.1 nodes at the second layer of dependency trees and peak at 237.8 affected nodes at the third layer. Notably, cascading risks are concentrated in critical hub nodes such as \texttt{transformers}, which directly or indirectly affect over 1,300 downstream packages. These findings provide quantitative insights into the structural and security dynamics of the LLMSC and emphasize the need for targeted mitigation strategies to enhance ecosystem resilience.

cs.SE↗