When Answer Decodability Is Not a Decision: Separating Final-Answer and Imminent-Emission Signals in Chain-of-Thought
cs.LG, cs.CL
Submitted: 2026-11-09
Updated: 2026-11-09
Terminology
Sources
- Understanding intermediate layers using linear classifier probes
- Chain-of-Thought Reasoning In The Wild Is Not Always Faithful
- The Internal State of an LLM Knows When It's Lying
- Probing Classifiers: Promises, Shortcomings, and Advances
- Eliciting Latent Predictions from Transformers with the Tuned Lens
- Reasoning Theater: Disentangling Model Beliefs from Chain-of-Thought
- Discovering Latent Knowledge in Language Models Without Supervision
- Post-Hoc Reasoning in Chain of Thought: Decoding and Steering Pre-Committed Answers
- Amnesic Probing: Behavioral Explanation with Amnesic Counterfactuals
- Causal Abstractions of Neural Networks
- Patchscopes: A Unifying Framework for Inspecting Hidden Representations of Language Models
- Measuring Massive Multitask Language Understanding
- Designing and Interpreting Probes with Control Tasks
- Towards Faithfully Interpretable NLP Systems: How should we define and evaluate faithfulness?
- Language Models (Mostly) Know What They Know
- Large Language Models are Zero-Shot Reasoners
- Measuring Faithfulness in Chain-of-Thought Reasoning
- The Chain Holds, the Answer Folds: Trace-Answer Dissociation in Reasoning Models Under Adversarial Pressure
- Let's Verify Step by Step
- When Chain-of-Thought Fails, the Solution Hides in the Hidden States
Related papers
- A Physics-Regulated Neural Framework for Learning 3D Grain Growth Dynamics
- FlexMoE: Learning One-for-All Multi-Budget Expert Pruning for Flexible MoE Deployment
- OptiMind: Improving Small Language Models for Optimization Modeling through Formulation Guidance and Multi-Turn Learning
- Staying in the Sweet Spot: Boost Reasoning Evolution via Responsive Rollout Difficulty Manipulation
- Examining the Impact of Data and Training Methods on Emergent and Subliminal Misalignment
- Rethinking Black-Box On-Policy Distillation: A Rubric-based Analysis