Daniel Fried

Researcher@Meta AI·美国·OpenReview

研究方向

code generation · LLMs · language grounding · vision and language · grounded instruction following · computational pragmatics · constituency parsing · syntactic parsing · structured prediction

Daniel Fried

Generative Value Conflicts Reveal LLM Priorities

From Reproduction to Replication: Evaluating Research Agents with Progressive Code Masking

MetaLint: Generalizable Idiomatic Code Quality Analysis Through Instruction-Following and Easy-to-Hard Generalization

BigCodeBench: Benchmarking Code Generation with Diverse Function Calls and Complex Instructions

Human-Aligned Chess With a Bit of Search

SWE-RL: Advancing LLM Reasoning via Reinforcement Learning on Open Software Evolution

RepoST: Scalable Repository-Level Coding Environment Construction with Sandbox Testing

Inducing Programmatic Skills for Agentic Tasks

Dissecting Adversarial Robustness of Multimodal LM Agents

Repetition Improves Language Model Embeddings

Tree Search for Language Model Agents

Agent Workflow Memory

CodeBenchGen: Creating Scalable Execution-based Code Generation Benchmarks

Agent Workflow Memory