dorsal/arxiv
View SchemaRISER: Orchestrating Latent Reasoning Skills for Adaptive Activation Steering
| Authors | Wencheng Ye, Liang Peng, Xiaoyang Yuan, Yi Bin, Pengpeng Zeng, Hengyu Jin, Heng Tao Shen |
|---|---|
| Categories | |
| ArXiv ID | 2601.09269vv2 |
| URL | https://arxiv.org/abs/2601.09269 |
| License | http://creativecommons.org/licenses/by-nc-nd/4.0/ |
Abstract
Recent work on domain-specific reasoning with large language models (LLMs) often relies on training-intensive approaches that require parameter updates. While activation steering has emerged as a parameter efficient alternative, existing methods apply static, manual interventions that fail to adapt to the dynamic nature of complex reasoning. To address this limitation, we propose RISER (Router-based Intervention for Steerable Enhancement of Reasoning), a plug-and-play intervention framework that adaptively steers LLM reasoning in activation space. RISER constructs a library of reusable reasoning vectors and employs a lightweight Router to dynamically compose them for each input. The Router is optimized via reinforcement learning under task-level rewards, activating latent cognitive primitives in an emergent and compositional manner. Across seven diverse benchmarks, RISER yields 3.4-6.5% average zero-shot accuracy improvements over the base model while surpassing CoT-style reasoning with 2-3x higher token efficiency and robust accuracy gains. Further analysis shows that RISER autonomously combines multiple vectors into interpretable, precise control strategies, pointing toward more controllable and efficient LLM reasoning.
{
"annotation_id": "78b31469-4afb-425b-b9be-8fb182c8b97e",
"date_created": "2026-02-17T05:53:20.221000Z",
"date_modified": "2026-02-17T05:53:20.221000Z",
"file_hash": "855b305d71fb5a323e75c7714069067445336cab48ec686ae92574b85c0ec7e3",
"private": false,
"record": {
"abstract": "Recent work on domain-specific reasoning with large language models (LLMs) often relies on training-intensive approaches that require parameter updates. While activation steering has emerged as a parameter efficient alternative, existing methods apply static, manual interventions that fail to adapt to the dynamic nature of complex reasoning. To address this limitation, we propose RISER (Router-based Intervention for Steerable Enhancement of Reasoning), a plug-and-play intervention framework that adaptively steers LLM reasoning in activation space. RISER constructs a library of reusable reasoning vectors and employs a lightweight Router to dynamically compose them for each input. The Router is optimized via reinforcement learning under task-level rewards, activating latent cognitive primitives in an emergent and compositional manner. Across seven diverse benchmarks, RISER yields 3.4-6.5% average zero-shot accuracy improvements over the base model while surpassing CoT-style reasoning with 2-3x higher token efficiency and robust accuracy gains. Further analysis shows that RISER autonomously combines multiple vectors into interpretable, precise control strategies, pointing toward more controllable and efficient LLM reasoning.",
"arxiv_id": "2601.09269",
"authors": [
"Wencheng Ye",
"Liang Peng",
"Xiaoyang Yuan",
"Yi Bin",
"Pengpeng Zeng",
"Hengyu Jin",
"Heng Tao Shen"
],
"categories": [
"cs.AI"
],
"license": "http://creativecommons.org/licenses/by-nc-nd/4.0/",
"title": "RISER: Orchestrating Latent Reasoning Skills for Adaptive Activation Steering",
"url": "https://arxiv.org/abs/2601.09269",
"version": "v2"
},
"schema_id": "dorsal/arxiv",
"source": {
"execution_id": "c5f7548c-63d0-44e3-b731-a83e482cd989",
"id": "arXiv Dataset",
"type": "Model",
"variant": "snapshot-2026-01-17",
"version": "0.1.0"
},
"user_id": 1000002
}