COMPASS: Finding Where Reasoning Lives in Language Models
cs.AI
Submitted: 2026-10-05
Updated: 2026-10-07
Code: https://github.com/pratyay2510/COMPASS
Sources
- Triggering Chain-of-Thought via Latent Feature Interventions in Large Language Models
- Implicit Chain of Thought Reasoning via Knowledge Distillation
- The Llama 3 Herd of Models
- Qwen3 Technical Report
- Gemma 4 Technical Report
- Training Verifiers to Solve Math Word Problems
- HARP: A challenging human-annotated math reasoning benchmark
- Fractional Reasoning via Latent Steering Vectors Improves Inference Time Compute
- Decomposing and Steering Functional Metacognition in Large Language Models
- Reasoning Models Know When They're Right: Probing Hidden States for Self-Verification
- Knowledge Knows, Verbalization Tells: Disentangling Latent Directions for Mathematical Solvability in LLMs
- Code Correctness Is Linearly Decodable from LLM Hidden States Before Generation
- Learning without training: The implicit dynamics of in-context learning
- Steering Language Models With Activation Engineering
- Representation Engineering: A Top-Down Approach to AI Transparency
- FineSteer: A Unified Framework for Fine-Grained Inference-Time Steering in Large Language Models
- Invariant Reasoning Directions in Latent Trajectories of Language Models
- The First to Know: How Token Distributions Reveal Hidden Knowledge in Large Vision-Language Models?
Related papers
- MAVEN-T: Reinforced Heterogeneous Distillation for Real-Time Multi-Agent Trajectory Prediction
- Model Discovery Agent: LLM-assisted Bayesian experiment design for data-efficient discovery of mechanistic world models
- Agents' Last Exam
- The Clinician's Veto: Navigating Trust, Liability, and Uncertainty in Autonomous AI Prescribing
- MindHelper: Closed-Loop Embodied Mental-State Reasoning for Precision Intervention
- Incumbent Advantage: Brand Bias and Cognitive Manipulation Dynamics in LLM Recommendation Systems