Chen
50 indexed papers
Publications per year
Top categories
Frequent co-authors
Research Timeline
This paper presents HiState, a prefix cache management system for hybrid models serving agent workloads to reduce inference costs.
A large language model-based chatbot improved patient visit preparation and satisfaction.
The paper introduces ProSearch, a benchmark for multi-constraint protocol retrieval in experimental science.
Develops a benchmark and language model to improve scientific reasoning in perovskite photovoltaics.
This paper explores the representation of forget data in vision-language model unlearning and introduces a modality routing strategy to improve unlearning effectiveness.
This paper proposes a new method, Feasibility-Preserving Transactional Refinement (FPTR), to improve the efficiency of short-interval wireless schedulers.
The paper introduces D3D, a standalone diffusion drafter that removes inter-round dependency between drafting and verification, enabling asynchronous speculative decoding.
The paper introduces FrontierPhysics, an end-to-end benchmark for evaluating AI agents in physics research.
The authors present HALO, a humanoid autonomy system that synthesizes vision-language-motion data to learn physical skills and solve composite tasks without human supervision.
WeLM-Image is a unified image foundation model for text-to-image generation, instruction-guided editing, and multi-reference synthesis.
The paper introduces XStack, a benchmark that evaluates agents on migrating ML repositories between PyTorch and JAX ecosystems.
PhysioWave-T is a token-efficient wavelet foundation model family for EEG, ECG, and sEMG.
The paper introduces Flex-RSI, a benchmark for evaluating the recursive self-improvement of agents on vision and robotics tasks.
This paper shows that generalist AI models can control manipulators more reliably than specialist policies.
The paper proposes GRAC, a framework that uses graph-based representations and retrieved examples to improve surgical scene captioning.
This paper proposes a flow-guided target-space alignment method called FlowAlign that improves instance-level correspondence while preserving target-space structure in multimodal alignment problems.
The paper studies leave-one-out attribution for language models and provides an efficient algorithm to estimate LOO values with a matching lower bound.
This paper presents Metal Sandbox, a system for efficient on-device inference on macOS that combines restricted file access, native command execution, and command-level scheduling.
The paper proposes AntEmbed, a dual-mode multimodal embedding model for efficient retrieval and reasoning-enhanced fine-grained matching.
The paper presents two results on the randomized communication complexity of composed functions, providing new composition lower bounds and strengthening previous results.