Qian Liu

Researcher@TikTok·新加坡·OpenReview

研究方向

multilingual language model · code generation · table pre-training · compositional generalization · conversational semantic parsing · semantic parsing

Qian Liu

SimpleTIR: End-to-End Reinforcement Learning for Multi-Turn Tool-Integrated Reasoning

TreePO: Enhancing Policy Efficacy and Inference Efficiency with Tree Modeling

DAComp: Benchmarking Data Agents across the Full Data Intelligence Lifecycle

SWE-Perf: Can Language Models Optimize Code Performance on Real-World Repositories?

SWE-Dev: Training and Evaluating Autonomous End-to-End Feature-Driven Software Development

BigCodeArena: Unveiling More Reliable Human Preferences in Code Generation via Execution

First return, entropy-eliciting explore

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

Spider 2.0: Evaluating Language Models on Real-World Enterprise Text-to-SQL Workflows

Cheating Automatic LLM Benchmarks: Null Models Achieve High Win Rates

When Attention Sink Emerges in Language Models: An Empirical View

Afterburner: Reinforcement Learning Facilitates Self-Improving Code Efficiency Optimization

SimpleRL-Zoo: Investigating and Taming Zero Reinforcement Learning for Open Base Models in the Wild

RegMix: Data Mixture as Regression for Language Model Pre-training

Improving Your Model Ranking on Chatbot Arena by Vote Rigging

Programming Every Example: Lifting Pre-training Data Quality Like Experts at Scale

ZeCO: Zero-Communication Overhead Sequence Parallelism for Linear Attention

Scaling up Masked Diffusion Models on Text

General-Reasoner: Advancing LLM Reasoning Across All Domains

SkyLadder: Better and Faster Pretraining via Context Window Scheduling

Programming Every Example: Lifting Pre-training Data Quality like Experts at Scale

Bootstrapping Language Models with DPO Implicit Rewards

Unnatural Languages Are Not Bugs but Features for LLMs

Predictive Data Selection: The Data That Predicts Is the Data That Teaches

Test-Time Backdoor Attacks on Multimodal Large Language Models