Research map / Large language models / Language, safety and interpretability
Polarity-Aware Probing for Quantifying Latent Alignment in Language Models
Sabrina Sadiekh, Elena Ericheva, Chirag Agarwal
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| paraphrase harmful/safe statements (concurrent vs antagonistic pairs) (ablation) | PA-CCS metrics on large models · Mixed vs Not datasets | PC and CI: MAD PC 0.030, CI 0.073 | No clear effect |
| remove the negation token not (random placeholder ttt control) (ablation) | PA-CCS metrics on large models · Not vs ttt datasets | PC and CI: MAD PC 0.274, CI 0.322 | Improved |
| use encoder instead of decoder models | PA-CCS probing · mixed, not, ttt datasets | variance of accuracy, PC, CI: median difference Accuracy 0.009, PC 0.016, CI 0.023 | Improved |
| add instruction tuning or task-specific fine-tuning | language models probed with PA-CCS · mixed, not, ttt datasets | accuracy, PC, CI | Improved |
| scale up model size (small <2B to large >=2B) | PA-CCS probing · mixed and not datasets | layers with separation accuracy >= 0.75: 1.6% -> 28% | Improved |
Closest papers by meaning
- LLM Alignment--Utility Asymmetry under Semantic-Preserving Transformations (NeurIPS 2026 · similarity 0.88)
- Exploring Intrinsic Alignments Within Text Corpus (AAAI 2025 · similarity 0.87)
- W2S-AlignTree: Weak-to-Strong Inference-Time Alignment for Large Language Models via Monte Carlo Tree Search (AAAI 2026 · similarity 0.87)
- Listening to the Wise Few: Query–Key Alignment Unlocks Latent Correct Answers in Large Language Models (NeurIPS 2026 · similarity 0.87)
- Monitoring Latent World States in Language Models with Propositional Probes (ICLR 2025 · similarity 0.87)
- PaCE: Parsimonious Concept Engineering for Large Language Models (NeurIPS 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
