RISER: Orchestrating Latent Reasoning Skills for Adaptive Activation Steering

2026-01-14

AI Summary

Existing activation intervention methods rely on static, manually designed strategies that struggle to adapt to the dynamic demands of complex reasoning tasks. This work proposes RISER, a novel framework that introduces the first reinforcement learning–based dynamic and composable activation intervention mechanism. By constructing a reusable library of reasoning vectors and a lightweight routing module, RISER adaptively composes activation signals to elicit latent cognitive primitives within the model—without updating its parameters. Evaluated across seven benchmarks, the method achieves zero-shot accuracy improvements of 3.4–6.5%, delivers 2–3× higher token efficiency compared to chain-of-thought approaches, and demonstrates consistently robust performance gains.

Technology Category

• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning
• Machine Learning: Active Learning
• Knowledge Representation and Reasoning: Common-Sense Reasoning

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

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.

Problem

• activation steering
• large language models
• reasoning
• adaptive intervention
• parameter efficiency

Innovation

• activation steering
• reasoning vectors
• reinforcement learning
• compositional reasoning
• parameter-efficient intervention