Skip to content
View nick7nlp's full-sized avatar
😀
I may be slow to respond.
😀
I may be slow to respond.
11 followers · 2 following

Achievements

Achievement: QuickdrawAchievement: Starstruck

Achievements

Achievement: QuickdrawAchievement: Starstruck

Block or report nick7nlp

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don't include any personal information such as legal names or email addresses. Markdown supported. This note will be visible to only you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
nick7nlp/README.md

     


📓 research notes

How do we make LLMs reason better without burning the planet?

Working on: curriculum RL, concise chain-of-thought, long-context evaluation

Recent: FastCuRL accepted at EMNLP 2025 ✨


📎 papers

  ☑️  Hunyuan-MT — WMT 2025 · 2025.09
        🥇 1st in 30/31 WMT25 language pairs · contributor

  ☑️  ConciseR — preprint · 2025.05
        walk before you run — concise reasoning via GRPO++ → L-GRPO

  ☑️  FastCuRL — EMNLP 2025 🎉 · 2025.03
        curriculum RL — SOTA with 50% fewer steps, 8 GPUs only

  ☑️  Counting-Stars — COLING 2025 · 2024.03
        multi-evidence, position-aware benchmark for long-context LLMs



previous work

Pinned Loading

  1. Tencent-Hunyuan/Hunyuan-MT Tencent-Hunyuan/Hunyuan-MT Public

    Python 683 67

  2. Counting-Stars Counting-Stars Public

    Counting-Stars (★)

    Jupyter Notebook 83 2

  3. FastCuRL FastCuRL Public

    FastCuRL: Curriculum Reinforcement Learning with Stage-wise Context Scaling for Efficient LLM Reasoning (EMNLP 2025)

    Python 57 4

  4. ConciseR ConciseR Public

    Walk Before You Run! Concise LLM Reasoning via Reinforcement Learning

    6