Same Answer, Different Future: Dissociating Belief Content and Evidence Weight in Language Models
cs.LG
Submitted: 2026-11-05
Updated: 2026-11-11
Terminology
Sources
- Large Language Models Develop Belief State Geometry In-Context
- LLMs are not (consistently) Bayesian: Quantifying internal (in)consistencies of LLMs' probabilistic beliefs
- Decomposing how prompting steers behavior
- LLMs are Bayesian, In Expectation, Not in Realization
- Verbalizable Representations Form a Global Workspace in Language Models
- Prompt-Activation Duality: Improving Activation Steering via Attention-Level Interventions
- Elicitation Matters: How Prompts and Query Protocols Shape LLM Surrogates under Sparse Observations
- Erased, Rerouted, or Rescaled? Post-Training and the Causal Quotient of a Language Model's Belief State
- Hidden APIs in Language Models: Discovering Reusable Causal Interfaces from Forked Futures
- The Shape of Beliefs: Geometry, Dynamics, and Interventions along Representation Manifolds of Language Models' Posteriors
- When Does Activation Steering Change What a Model Computes From?
- Believe It or Not: How Deeply do LLMs Believe Implanted Facts?
- Beliefs and Behavior in Language Models
- Resist and Update: Counterfactual Report Coordinates for Incentive-Compatible LLMs
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