Minimally Invasive Steering of Language Models
cs.LG, cs.AI
Submitted: 2026-09-24
Updated: 2026-09-24
Terminology
Sources
- Concept Heterogeneity-aware Representation Steering
- LFM2 Technical Report
- Constitutional AI: Harmlessness from AI Feedback
- KTO: Model Alignment as Prospect Theoretic Optimization
- Scaling Laws for Reward Model Overoptimization
- SimCSE: Simple Contrastive Learning of Sentence Embeddings
- Aligning Language Models with Preferences through f-divergence Minimization
- Test-time reward-guided alignment of language models by importance sampling on pre-logit space
- Preemptive Detection and Steering of LLM Misalignment via Latent Reachability
- Skywork-Reward-V2: Scaling Preference Data Curation via Human-AI Synergy
- Self-Refine: Iterative Refinement with Self-Feedback
- Many-Tier Instruction Hierarchy in LLM Agents
- Steering Language Models With Activation Engineering
- Self-Consistency Improves Chain of Thought Reasoning in Language Models
- GenARM: Reward Guided Generation with Autoregressive Reward Model for Test-time Alignment
Related papers
- Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
- AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
- Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
- Convergence issues in Relational Concept Analysis based on AOC-posets
- Beliefs Beyond Posteriors: Local-Consistency Optimisation for Bayesian Neural Networks
- Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks