Qwen2.5-1.5B-Instruct-Distill-GRPO-Step80-LogicRL

affa/Qwen2.5-1.5B-Instruct-Distill-GRPO-Step80-LogicRL

Qwen2.5-1.5B-Instruct-Distill-GRPO-Step80-LogicRL is a machine learning model from affa. Use it for the machine learning task on the model card, and read the license before you ship it in a product. The card lists the license as mit.

We applied GRPO to the distilled model, resulting in stable training and performance comparable to the 7B model.

Downloads · 30 days

24% of all-time downloads

All-time downloads

Public

Parameters

1.8B

3.6 GB on disk

Likes

Public

Hugging Face

Repo makeup

Click a slice to open those files.

.safetensors3.6 GB · 100%

At a glance

License: mit
Model type: qwen2
Access: Public
Created: Aug 27, 2025
Updated: Aug 27, 2025
SHA: 7c3f0242

Model card

From the Hugging Face model README

Cold Starting RL on the Distilled Model

We applied GRPO to the distilled model, resulting in stable training and performance comparable to the 7B model. image/png

Model	Avg	2ppl	3ppl	4ppl	5ppl	6ppl	7ppl	8ppl
o3-mini-high	0.94	0.99	0.98	0.97	0.95	0.94	0.89	0.83
o1-2024-12-17	0.42	0.83	0.51	0.38	0.38	0.35	0.30	0.20
GPT-4o	0.37	0.68	0.57	0.49	0.32	0.23	0.21	0.11
Qwen2.5-7B-Instruct-1M	0.26	0.64	0.39	0.33	0.21	0.13	0.03	0.08
Qwen2.5-1.5B-Instruct-Distill-GRPO-Step80-LogicRL	0.80	0.99	0.97	0.98	0.91	0.74	0.55	0.52

Type: qwen2
License: mit
Created: Aug 27, 2025
Updated: Aug 27, 2025