Search results

GRPO training with Qwen2.5-1.5B-Instruct leads to performance drop below base model

Status: Open.

#21 In knoveleng/open-rs;

· naajeehxe opened on Mar 19, 2026

Is there a script or guide for SFT training?

Status: Open.

#20 In knoveleng/open-rs;

· naajeehxe opened on Mar 18, 2026

Can this project be adapted to train Gemma models instead of Qwen?

Status: Open.

#18 In knoveleng/open-rs;

· seqseq opened on Sep 3, 2025

How was the dataset used in the experiment filtered out

Status: Open.

#16 In knoveleng/open-rs;

· programmerIm opened on Jun 15, 2025

Questions about LightEval Results and Baseline Accuracy in Your Paper

Status: Open.

#15 In knoveleng/open-rs;

· Kairong-Han opened on May 27, 2025

Can't reproduce the result of Open-RS3, which is reported in paper?

Status: Open.

#12 In knoveleng/open-rs;

· xiaozhang23957 opened on Apr 1, 2025