Downloads · 30 days
5
23% of all-time downloads
l1-max-qwen3.5-9b-seven-task
wckwan/l1-max-qwen3.5-9b-seven-task
l1-max-qwen3.5-9b-seven-task is a machine learning model from wckwan. 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 apache-2.0.
Qwen/Qwen3.5-9B trained with grpo on the seven-task setting: 50 steps, 32 prompts x 8 rollouts per step, lr 2e-6, KL 1e-3, 32K rollout cap, verl v0.9.1 (FSDP2 + vLLM). Training data: 512 examples (seed 0) from the off…
All-time downloads
22
Public
Parameters
9.4B
37.7 GB on disk
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors37.6 GB · 100%
At a glance
- License
- apache-2.0
- Model type
- qwen3_5
- Access
- Public
- Created
- Sep 21, 2026
- Updated
- Sep 22, 2026
- SHA
- 05bd6c4c
Base models
Model card
From the Hugging Face model README
l1-max-qwen3.5-9b-seven-task
Qwen/Qwen3.5-9B trained with grpo on the seven-task setting: 50 steps, 32 prompts x 8 rollouts per step, lr 2e-6, KL 1e-3, 32K rollout cap, verl v0.9.1 (FSDP2 + vLLM). Training data: 512 examples (seed 0) from the official training splits; system prompt "Solve the user's task and give the final answer directly."; reward = task correctness of the text after </think>.
L1 max stage (Aggarwal & Welleck, 2025): length-controlled reward thinkgram/rl/l1_reward.py, alpha 0.0003, budgets 64-4096 tokens (max mode up to 32768), started from /dev/volume/checkpoints/structured-thinking/l1-exact-qwen3.5-9b-seven-task/global_step_50/actor/huggingface.
- Type
- qwen3_5
- License
- apache-2.0
- Created
- Sep 21, 2026
- Updated
- Sep 22, 2026