Skip to content

@thunlp THUNLP

Change the repository type filter

All

    Repositories list

    264 repositories

    • SE-Bench

      Public
      Official repo for "SE-Bench: Benchmarking Self-Evolution with Knowledge Internalization"
      Python
      •
      MIT License
      •3•19•4•0•Updated Feb 24, 2026Feb 24, 2026

    • JustRL

      Public
      [ICLR 2026 Blogpost Track Poster] JustRL: Scaling a 1.5B LLM with a Simple RL Recipe
      Python
      •10•250•1•0•Updated Feb 23, 2026Feb 23, 2026

    • NOSA

      Public
      The official implementation of NOSA
      Python
      •0•15•0•0•Updated Feb 12, 2026Feb 12, 2026

    • ProactiveAgent

      Public
      A LLM-based Agent that predict its tasks proactively.
      Python
      •
      Apache License 2.0
      •39•499•5•0•Updated Feb 10, 2026Feb 10, 2026

    • hybrid-linear-attention

      Public
      Code and models for the paper: Hybrid Linear Attention Done Right: Efficient Distillation and Effective Architectures for Extremely Long Contexts
      Python
      •2•24•1•0•Updated Feb 4, 2026Feb 4, 2026

    • APB

      Public
      Official Implementation of APB (ACL 2025 main Oral) and Spava.
      C++
      •4•34•0•0•Updated Jan 30, 2026Jan 30, 2026

    • ACDiT

      Public
      ACDiT: Interpolating Autoregressive Conditional Modeling and Diffusion Transformer
      Python
      •
      MIT License
      •1•41•2•0•Updated Jan 29, 2026Jan 29, 2026

    • KG-Infused-RAG

      Public
      Official implementation for the paper "KG-Infused RAG: Augmenting Corpus-Based RAG with External Knowledge Graphs"
      Python
      •2•21•0•0•Updated Jan 18, 2026Jan 18, 2026

    • H-Neurons

      Public
      The official implementation of the paper: H-Neurons: On the Existence, Impact, and Origin of Hallucination-Associated Neurons in LLMs
      Python
      •
      MIT License
      •0•21•0•0•Updated Jan 14, 2026Jan 14, 2026

    • BlockFFN

      Public
      Source codes for paper "BlockFFN: Towards End-Side Acceleration-Friendly Mixture-of-Experts with Chunk-Level Activation Sparsity".
      Python
      •5•18•0•0•Updated Jan 10, 2026Jan 10, 2026

    • LLaVA-UHD

      Public
      LLaVA-UHD v3: Progressive Visual Compression for Efficient Native-Resolution Encoding in MLLMs
      Python
      •
      Apache License 2.0
      •21•415•7•0•Updated Dec 20, 2025Dec 20, 2025

    • ChartCoder

      Public
      [ACL'25 Main] ChartCoder: Advancing Multimodal Large Language Model for Chart-to-Code Generation
      Python
      •4•77•2•0•Updated Dec 8, 2025Dec 8, 2025

    • LLMxMapReduce

      Public
      Python
      •
      Apache License 2.0
      •62•863•0•0•Updated Nov 6, 2025Nov 6, 2025

    • StateX

      Public
      The official implementation of the paper "StateX: Enhancing RNN Recall via Post-training State Expansion".
      machine-learningmemoryrnnrecallssmmambalinear-attentionllmlong-context
      Python
      •0•3•0•0•Updated Oct 24, 2025Oct 24, 2025

    • AgentRM

      Public
      [ACL 2025 main] AgentRM: Enhancing Agent Generalization with Reward Modeling
      Python
      •0•6•0•0•Updated Sep 29, 2025Sep 29, 2025

    • stuffed-mamba

      Public
      The code of the paper Stuffed Mamba: Oversized States Lead to the Inability to Forget
      machine-learningrnnmambalong-context
      Python
      •0•1•0•0•Updated Sep 28, 2025Sep 28, 2025

    • BurstEngine

      Public
      BurstEngine is an efficient framework designed to train LLMs on long-sequence data.
      Python
      •3•9•0•0•Updated Sep 25, 2025Sep 25, 2025

    • cost-optimal-gqa

      Public
      The code for the paper "Cost-Optimal Grouped-Query Attention for Long-Context Modeling"
      natural-language-processingtransformerattentionlong-contextllms
      Python
      •1•4•1•0•Updated Sep 14, 2025Sep 14, 2025

    • SIR-Bench

      Public
      Python
      •
      Apache License 2.0
      •0•5•1•0•Updated Sep 12, 2025Sep 12, 2025

    • Seq1F1B

      Public
      Sequence-level 1F1B schedule for LLMs.
      Python
      •
      Other
      •3.6k•38•1•0•Updated Aug 26, 2025Aug 26, 2025

    • FR-Spec

      Public
      [ACL 2025 main] FR-Spec: Frequency-Ranked Speculative Sampling
      C++
      •2•51•3•0•Updated Jul 15, 2025Jul 15, 2025

    • TritonBench

      Public
      TritonBench: Benchmarking Large Language Model Capabilities for Generating Triton Operators
      Python
      •
      Apache License 2.0
      •13•115•4•1•Updated Jun 14, 2025Jun 14, 2025

    • ClueAnchor

      Public
      Python
      •0•9•0•0•Updated Jun 11, 2025Jun 11, 2025

    • DeepPerception

      Public
      DeepPerception: Advancing R1-like Cognitive Visual Perception in MLLMs for Knowledge-Intensive Visual Grounding
      Python
      •
      MIT License
      •1•66•1•0•Updated Jun 10, 2025Jun 10, 2025

    • TAADpapers

      Public
      Must-read Papers on Textual Adversarial Attack and Defense
      nlpnatural-language-processingadversarial-learningadversarial-attackspaper-listadversarial-defense
      Python
      •
      MIT License
      •194•1.6k•3•0•Updated Jun 4, 2025Jun 4, 2025

    • LongPiBench

      Public
      Python
      •
      MIT License
      •0•0•0•0•Updated May 28, 2025May 28, 2025

    • DIET

      Public
      Official code for "The Overthinker's DIET: Cutting Token Calories with DIfficulty-AwarE Training"
      Python
      •0•1•0•0•Updated May 27, 2025May 27, 2025

    • Migician

      Public
      [ACL2025 Findings] Migician: Revealing the Magic of Free-Form Multi-Image Grounding in Multimodal Large Language Models
      Python
      •
      MIT License
      •4•89•1•0•Updated May 20, 2025May 20, 2025

    • ToLeaP

      Public
      Python
      •
      MIT License
      •1•5•1•1•Updated May 17, 2025May 17, 2025

    • SICOG

      Public
      Will Pre-Training Ever End? A First Step Toward Next-Generation Foundation MLLMs via Self-Improving Systematic Cognition
      Python
      •
      GNU General Public License v3.0
      •2•31•1•0•Updated May 14, 2025May 14, 2025