Downloads · 30 days
10
24% of all-time downloads
agentgym-rl-sciworld-qwen2.5-7b-plan-k3-nognorm
PolarisDane/agentgym-rl-sciworld-qwen2.5-7b-plan-k3-nognorm
agentgym-rl-sciworld-qwen2.5-7b-plan-k3-nognorm is a reinforcement learning model from PolarisDane. Use it for the reinforcement learning task on the model card, and read the license before you ship it in a product. It is set up for transformers. The card lists the license as apache-2.0.
ScienceWorld, Qwen2.5-7B-Instruct, plan forecast K=3（不做 groupnorm）
All-time downloads
42
Public
Parameters
7.6B
15.6 GB on disk
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors15.2 GB · 100%
At a glance
- Task
- Reinforcement Learning
- Library
- transformers
- License
- apache-2.0
- Model type
- qwen2
- Access
- Public
- Created
- Sep 21, 2026
- Updated
- Sep 21, 2026
- SHA
- f9134fa1
Model card
From the Hugging Face model README
agentgym-rl-sciworld-qwen2.5-7b-plan-k3-nognorm
ScienceWorld, Qwen2.5-7B-Instruct, plan forecast K=3（不做 group_norm）
用 AgentGym-RL 框架训练的 agentic RL 检查点。
评测
| step | 成功率 | Average Score |
|---|---|---|
| 250 | 72.50% | 53.40 |
成功率口径为 done 且 score = 100（严格：任务真正完成），200 题官方 eval split，max_rounds=30, temperature=1。
检查点
每个训练 step 是一个独立分支，main 指向 step 250。
step-250← main
from transformers import AutoModelForCausalLM, AutoTokenizer
m = AutoModelForCausalLM.from_pretrained("PolarisDane/agentgym-rl-sciworld-qwen2.5-7b-plan-k3-nognorm", revision="step-250")
t = AutoTokenizer.from_pretrained("PolarisDane/agentgym-rl-sciworld-qwen2.5-7b-plan-k3-nognorm", revision="step-250")
- Task
- Reinforcement Learning
- Library
- transformers
- Type
- qwen2
- License
- apache-2.0
- Created
- Sep 21, 2026
- Updated
- Sep 21, 2026