Resources

Research Directions

Multilingual Translation, Generation and Alignment

Low-resource Languages, Speech-to-Text Translation, LLM Evaluation, Alignment

Hire a Linguist!: Learning Endangered Languages with In-Context Linguistic Descriptions

LLMRefine: Pinpointing and Refining Large Language Models via Fine-Grained Actionable Feedback

Pride and Prejudice: LLM Amplifies Self-Bias in Self-Refinement

INSTRUCTSCORE: Explainable Text Generation Evaluation with Finegrained Feedback

WACO: Word-Aligned Contrastive Learning for Speech Translation

Trustworthy Generative AI

Safety, Security, Privacy, and Copyright in GenAI/LLM

Permute-and-Flip: An Optimally Robust and Watermarkable Decoder for LLMs

Provable Robust Watermarking for AI-Generated Text

Invisible Image Watermarks Are Provably Removable Using Generative AI

Protecting Language Generation Models via Invisible Watermarking

Provably Confidential Language Modelling

Agentic Large Language Models

Code Generation, Reasoning and Planning

ALGO: Synthesizing Algorithmic Programs with LLM-Generated Oracle Verifiers

Don't Fine-Tune, Decode: Syntax Error-Free Tool Use via Constrained Decoding

Learning from Mistakes via Cooperative Study Assistant for Large Language Models

AutoPlan: Automatic Planning of Interactive Decision-Making Tasks With Large Language Models

AI-powered Drug Design

Generative models for Small-molecule and Protein Design, Molecule understanding.

Generative Enzyme Design Guided by Functionally Important Sites and Small-Molecule Substrates

SurfPro: Functional Protein Design Based on Continuous Surface

Importance Weighted Expectation-Maximization for Protein Sequence Design

Accelerating Antimicrobial Peptide Discovery with Latent Structure