Research map / Large language models / Language, safety and interpretability
Is This Just Fantasy? Language Model Representations Reflect Human Judgments of Event Plausibility
Michael A. Lepori, Jennifer Hu, Ishita Dasgupta, Roma Patel, Thomas Serre, Ellie Pavlick
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 |
|---|---|---|---|
| classify with modal difference vectors instead of sentence probability | LMs with >=2B parameters · Modal categorization generalization datasets | classification accuracy | Improved |
| use modal difference vectors in models with <2B parameters | GPT2, Llama-3.2-1B, OLMo-2-1B · Modal categorization generalization datasets | classification accuracy | Mixed |
| featurize with modal difference vector projections instead of baselines for human response modeling | logistic regression on LM features · Human modal categorization data | correlation / MSE | Improved |
| keep all three modal difference features (vs ablating one) (ablation) | human categorization regression · four-category human dataset | overall correlation and MSE | Improved |
| keep all three features (vs ablating one) (ablation) | human categorization regression · binary-judgment human datasets | overall correlation and MSE | No clear effect |
| steer generation with modal difference vectors (scale 5 vs 3) (ablation) | Gemma-2-2B, Llama-3.2-3B · 30 prefixes | surprisal of generated tokens | Improved |
Closest papers by meaning
- Potemkin Understanding in Large Language Models (ICML 2025 · similarity 0.86)
- Same Task, Different Circuits: Disentangling Modality-Specific Mechanisms in VLMs (NeurIPS 2025 · similarity 0.86)
- The Grounding Gap: How LLMs Anchor the Meaning of Abstract Concepts Differently from Humans (NeurIPS 2026 · similarity 0.86)
- Cross-modal Associations in Vision and Language Models: Revisiting the Bouba-Kiki Effect (NeurIPS 2025 · similarity 0.86)
- Less is More: Local Intrinsic Dimensions of Contextual Language Models (NeurIPS 2025 · similarity 0.86)
- XModBench: Benchmarking Cross-Modal Capabilities and Consistency in Omni-Language Models (ICLR 2026 · 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.
