Wen-tau Yih

Research Scientist@Meta Platforms, Inc.·美国·OpenReview

研究方向

Question Answering · Semantic Parsing · Neural Retriever · Retrieval Augmented Language Models

5.0

Learning Facts at Scale with Active Reading

ICLR 2026Poster

5.0

SCRIBES: Web-Scale Script-Based Semi-Structured Data Extraction with Reinforcement Learning

ICLR 2026Poster

4.5

HybridCoT: Interleaving Latent and Text Chain-of-Thought for Efficient Reasoning

ICLR 2026Rejected

通讯

4.5

SelfCite: Self-Supervised Alignment for Context Attribution in Large Language Models

合作者 (20)

Vincent-Pierre Berges

Wen-tau Yih

Learning Facts at Scale with Active Reading

SCRIBES: Web-Scale Script-Based Semi-Structured Data Extraction with Reinforcement Learning

HybridCoT: Interleaving Latent and Text Chain-of-Thought for Efficient Reasoning

TruthRL: Incentivizing Truthful LLMs via Reinforcement Learning

Learning to Reason for Factuality

CRAG-MM: Multi-modal Multi-turn Comprehensive RAG Benchmark

Semi-parametric language model with selective memory

ConfRAG: Confidence-Guided Retrieval-Augmenting Generation

Continual Learning via Sparse Memory Finetuning

FlexOLMo: Open Language Models for Flexible Data Use

Meta CLIP 2: A Worldwide Scaling Recipe

Memory Layers at Scale

ReasonIR: Training Retrievers for Reasoning Tasks

Group-Level Data Selection for Efficient Pretraining

SelfCite: Self-Supervised Alignment for Context Attribution in Large Language Models