---
license: apache-2.0
library_name: transformers
pipeline_tag: text-generation
tags:
- qwen3
- reinforcement-learning
- grpo
---

# qwen3-4b-material-grpo

Fine-tuned from `Qwen/Qwen3-4B` with GRPO on the `material` split.

## Validation Performance

Metric: `val-aux/sciknoweval/reward/mean@16` from 10-step validation logs.

| best mean@16 | best step | final mean@16 | final step |
|---:|---:|---:|---:|
| 78.32% | 40 | 66.16% | 100 |

![Validation mean@16](eval_mean16.png)

| step | mean@16 |
|---:|---:|
| 10 | 67.75% |
| 20 | 72.54% |
| 30 | 77.13% |
| 40 | 78.32% |
| 50 | 58.78% |
| 60 | 61.64% |
| 70 | 70.01% |
| 80 | 66.82% |
| 90 | 67.89% |
| 100 | 66.16% |

Files included with this repo:
- `metrics.json`: parsed validation summary
- `eval_mean16.csv`: step-level validation curve data
- `eval_mean16.png`: validation curve plot

Important: the uploaded weights are the final `global_step_100/actor` checkpoint. If `best step` is earlier than 100, the best validation point is reported for tracking, but the corresponding actor weights may not be retained locally.

Checkpoint source:
`/mnt/mole/SDPO/L2T/checkpoints/datasets/sciknoweval/material/qwen3gen-material-GRPO-Qwen-Qwen3-4B-mbs8-train64-rollout8-lr1e-6-vllm0.8`

W&B run: `run-20260630_003605-tfw9z9dp`

This upload uses `global_step_100/actor` converted from VERL FSDP shards to Hugging Face format.