---
license: other
license_name: qwen
license_link: https://huggingface.co/Qwen/Qwen3.5-9B/blob/main/LICENSE
base_model: Qwen/Qwen3.5-9B
pipeline_tag: text-generation
library_name: transformers
tags:
  - agent
  - tool-use
  - skill-selection
  - reinforcement-learning
  - grpo
---

# SkillGate-9B

Policy from **"SkillGate: Training In-Policy Skill Selection in Long-Horizon Agents."**

Agent frameworks expose *skills* — instruction files with a name, a one-line
description and a body — by progressive disclosure: the agent sees only names and
descriptions and must open a file to learn what is inside. With thousands of skills in
a library, *which* one to read becomes a decision the policy makes mid-episode, and
outcome-rewarded RL cannot teach it: the tokens naming the chosen skill carry a median
0.14% of their trajectory's loss weight, and two in five of them receive a *negative*
advantage because execution afterwards failed.

SkillGate partitions one trajectory's token support into two disjoint credit channels:
outcome credit reaches only execution tokens (the whole skill-read call is removed from
the task loss), while an action-local advantage reaches exactly the skill-naming
tokens, positive only when the trajectory's single read is the correct skill.

## Model

| | |
|---|---|
| Base | Qwen3.5-9B |
| Training | 100 steps on-policy GRPO, 491 tasks, 8 rollouts/prompt, global batch 128, lr 1e-6, KL 3e-5, selector coefficient 0.20 |
| Checkpoint | `iter_0000099`, the final step (`selection_role: final`) |
| Architecture | `Qwen3_5ForConditionalGeneration` |

## Results (385-trial protocol, 5 agentic benchmarks, 16-candidate slate)

| Method | Overall | Oracle read | Misleading read |
|---|---:|---:|---:|
| SFT (RL init) | 40.8 | 37.9 | 61.8 |
| SkillRL (outcome reward only) | 47.0 | 54.3 | 69.6 |
| **SkillGate** | **53.2** | **83.9** | **21.8** |

Same initialisation, data, steps and hyperparameters as the outcome-only row; the only
difference is which tokens the gradient reaches.

## Intended use

Research on agentic skill/tool selection. The model expects the OpenClaw-style prompt
profile and tool schema used in the paper; see the repository for the exact system
prompt and the frozen skill slates.

## Links

- Paper: [arXiv:2608.18852](https://arxiv.org/abs/2608.18852)
- Code: https://github.com/DeepExperience/SkillGate

## License

Derived from Qwen3.5-9B and distributed under the Qwen license; see `license_link`.