Issues

• Arena-Hard-V2 judge: an unrecognized/garbled verdict silently scores as a tie and is never flagged failed

  microsoft/LMOps#445 · Filter by author shaurya416 opened on Sep 24, 2026

• [OEL] Table 2 "In-Context": K=10 aggregation, act-budget truncation, and a possible budget-dependent reversal of the self-vs-cross ordering

  microsoft/LMOps#441 · Filter by author Estrellajer opened on Aug 5, 2026

• MiniLLM : Results on test split between evaluate() in finetune.py and evaluate_main() in evaluate_main.py are different

  microsoft/LMOps#418 · Filter by author sofianeElguendouz opened on May 20, 2026
  1

• Can the current black-box GAD addition be directly used in Qwen 3.5?

  microsoft/LMOps#402 · Filter by author han2-l opened on May 7, 2026
  2

  

• 【GAD】使用Qwen3-VL进行GAD训练极慢

  microsoft/LMOps#354 · Filter by author yph22 opened on Jan 29, 2026
  2

  

• [GAD] Question regarding Reward Score fluctuations and downstream evaluation

  microsoft/LMOps#353 · Filter by author ZhangChi315 opened on Jan 26, 2026
  2

  

• [MiniLLM] baseline hyperparameter settings

  microsoft/LMOps#333 · Filter by author hyhy892 opened on Nov 10, 2025
  1

• [MiniLLM] The download link for OpenWebText is invalid.

  microsoft/LMOps#332 · Filter by author hyhy892 opened on Nov 4, 2025
  3

• [MiniLLM] Strange Rouge-L result on Dolly dataset

  microsoft/LMOps#329 · Filter by author Stephen-K1 opened on Sep 11, 2025
  2

• [MiniLLM] Why the teacher prob are not extract offline?

  microsoft/LMOps#328 · Filter by author huge123 opened on Aug 9, 2025
  2

• the wrong of the ProTeGi

  microsoft/LMOps#327 · Filter by author cccchou opened on Jul 31, 2025

• 我尝试用论文《DATA SELECTION VIA OPTIMAL CONTROL FOR LANGUAGE MODELS》方法去筛选带有错误标签的数据，发现错误得标签也有可能会有较高得得分。

  microsoft/LMOps#326 · Filter by author huangluyao opened on Jul 24, 2025
  1

• ProTeGi：When using UCB evaluation, how many eval_rounds are set?

  microsoft/LMOps#325 · Filter by author qiuxiaoyu9954 opened on Jul 10, 2025

• [ProTeGi] python environment

  microsoft/LMOps#323 · Filter by author lbuera opened on Jun 10, 2025

• [MiniLLM] Questions about performance without LM pipeline

  microsoft/LMOps#322 · Filter by author pipilia opened on Jun 8, 2025

• [MiniLLM] roberta dataset differences between llama version and opt version

  microsoft/LMOps#321 · Filter by author pipilia opened on Jun 4, 2025
  4

• Release GeAR artifacts (models, dataset) on Hugging Face

  microsoft/LMOps#320 · Filter by author NielsRogge opened on Jun 2, 2025

• [minillm] does qwen2 support model parallelism?

  microsoft/LMOps#318 · Filter by author 01yingxuan opened on May 29, 2025
  3

• SFT validation loss of minillm

  microsoft/LMOps#317 · Filter by author lucky1day opened on May 26, 2025
  1

• [instruction-pretrain] Prompt customization for instruction-response generation

  microsoft/LMOps#313 · Filter by author ngogiahan149 opened on May 5, 2025
  1

• bug in teacher lora loading

  microsoft/LMOps#306 · Filter by author water1981929 opened on Apr 5, 2025

• MiniLLM LLaMA-Torch Shape Mismatch for Model Parallel

  microsoft/LMOps#305 · Filter by author majidurrahman1437 opened on Mar 16, 2025
  8

• [ProTeGi]where 's the file ethos_ishate_binary_shuf.csv?

  microsoft/LMOps#304 · Filter by author Thaumaturge2020 opened on Mar 16, 2025
  2

• DPKD ： A simple improvement of MiniLLM using DPO

  microsoft/LMOps#302 · Filter by author chuangzhidan opened on Mar 8, 2025

• [understand_icl] Why is W_ICL derived from the error signal generated in the demonstrations?

  microsoft/LMOps#301 · Filter by author jiogenes opened on Feb 27, 2025