Yimin

The Implicit Drifting Policy (IDP) is a novel one-step action generation framework that implicitly enforces trajectory correction constraints by analyzing local expert action geometry, overcoming the difficulties of explicitly estimating a training-time drifting field.

The paper introduces Med-HEAL, a comprehensive framework and dataset for systematically identifying and mitigating hallucinations in medical LLMs, demonstrating that a self-critique pipeline significantly improves model accuracy.

The paper proposes SimSD, a plug-and-play speculative decoding algorithm that adapts diffusion language models (dLLMs) to achieve fast, token-level acceleration by restoring causal masking capabilities.

The paper proposes FLAME, a novel framework that detects AI-generated image forgeries by identifying intrinsic energy anomalies caused by the diffusion process, achieving state-of-the-art localization.

The paper proposes Credit-Attenuated Privileged Feedback (CAPF), a training-time mechanism that uses verifier-side information to guide LLM search agents, significantly improving their performance on complex QA tasks.

The paper argues that current embodied planning benchmarks prioritize superficial language prediction over true physical reasoning, introducing new benchmarks and a large-scale dataset to demonstrate that physically grounded causal reasoning is necessary for reliable autonomous agents.

This paper introduces Imaginative Perception Tokens (IPT) to improve spatial reasoning in vision language models.

The paper introduces ReproRepo, a scalable framework for evaluating the reproducibility of machine learning research using LLM agents and human-raised GitHub issues.

DigenRL is a disaggregated RL framework for diffusion-based generative LLMs that achieves 1.56-2.10x throughput improvements over state-of-the-art diffusion RL systems.

This paper introduces DiStash, a disaggregated transactional key-value store that enables an application to use a single transaction to manage key-value pairs across different pools of stashes, preventing race conditions and data loss.

This paper proposes Joint Speech-Text Interleaved Pretraining (JSTIP) for speech recognition, which constructs interleaved speech-text sequences and achieves consistent entity accuracy improvement.

The paper introduces MedPMC, a framework that transforms permissively licensed literature into high-fidelity infrastructure for medical multimodal models, resulting in improved performance on various benchmarks.

This paper introduces DiPhon, a diffusion framework for size-scalable graph generation, using a continuous diffusion process on the graphon space and a discretized graph-level process.

This paper introduces LingBot-VA 2.0, a video-action foundation model designed for embodiment, with semantic visual-action tokenization, causal pretraining, sparse MoE backbone, and enhanced asynchronous inference.

The paper introduces PS4, a framework for training target speaker extraction models using a large-scale corpus and proxy-supervised joint training strategy.

This paper presents the benefits of visual pretraining for foundation model intelligence, outperforming text-only pretraining on multiple backbones and benchmarks.

The paper proposes SAGA, a framework for schema-aware grounding in agentic text-to-SPARQL generation, which maintains a persistent type state, filters incompatible property candidates, and handles missing schema information permissively.

xDSM is a full-space, elastic DSM system built over CXL that transparently scales unmodified multithreaded applications by employing an OS-runtime co-design, dynamic data placement policy, and spatial locality-aware elasticity.

This paper develops FGDSE, a feature-governed dynamic stacking ensemble for climate-resilient charging-asset management in electric vehicles, which predicts daily fault risk over a multi-week horizon and identifies extreme heat as a causal factor for heat-sensitive posts.

Specula is an autonomous system that generates high-quality formal specifications for large, complex code using LLMs, improving understanding and finding bugs.