Search results

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

Status: Open.

#418 In microsoft/LMOps;

· sofianeElguendouz opened on May 20, 2026

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

Status: Open.

#402 In microsoft/LMOps;

· han2-l opened on May 7, 2026

【GAD】使用Qwen3-VL进行GAD训练极慢

Status: Open.

#354 In microsoft/LMOps;

· yph22 opened on Jan 29, 2026

[GAD] Question regarding Reward Score fluctuations and downstream evaluation

Status: Open.

#353 In microsoft/LMOps;

· ZhangChi315 opened on Jan 26, 2026

[MiniLLM] baseline hyperparameter settings

Status: Open.

#333 In microsoft/LMOps;

· hyhy892 opened on Nov 10, 2025

[MiniLLM] The download link for OpenWebText is invalid.

Status: Open.

#332 In microsoft/LMOps;

· hyhy892 opened on Nov 4, 2025

[MiniLLM] Strange Rouge-L result on Dolly dataset

Status: Open.

#329 In microsoft/LMOps;

· Stephen-K1 opened on Sep 11, 2025

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

Status: Open.

#328 In microsoft/LMOps;

· huge123 opened on Aug 9, 2025

the wrong of the ProTeGi

Status: Open.

#327 In microsoft/LMOps;

· 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。

Status: Open.

#326 In microsoft/LMOps;

· huangluyao opened on Jul 24, 2025

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

Status: Open.

#325 In microsoft/LMOps;

· qiuxiaoyu9954 opened on Jul 10, 2025

[ProTeGi] python environment

Status: Open.

#323 In microsoft/LMOps;

· lbuera opened on Jun 10, 2025