Research map / Large language models / AI and society
Learning to Theorize the World from Observation
Doojin Baek, Gyubin Lee, Junyeob Baek, Hosung Lee, Sungjin Ahn
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 |
|---|---|---|---|
| add state grounding loss (ablation) | NEO · GridWorld alpha=0.33 Comp. OOD | transferability: 0.000 -> 0.933 | Improved |
| add state grounding loss (ablation) | NEO · GridWorld alpha=0.33 In-distribution | transferability: 0.000 -> 0.911 | Improved |
| increase codebook size to 36 (ablation) | NEO · GridWorld alpha=0.33 Comp. OOD | transferability: 0.933 -> 0.976 | Improved |
| increase codebook size to 36 (ablation) | NEO · GridWorld alpha=0.33 Length OOD | transferability: 0.845 -> 0.930 | Improved |
| raise lambda_MDL to 1.2 (ablation) | NEO · GridWorld alpha=0.33 Comp. OOD | transferability: 0.933 -> 0.160 | Hurt |
| lower lambda_MDL to 0.8 (ablation) | NEO · GridWorld alpha=0.33 Comp. OOD | transferability: 0.933 -> 0.956 | Mixed |
| add test-time search with majority voting (NEO-S) | NEO · Arithmetic alpha=0.33 Length OOD | transferability: 0.038 -> 0.524 | Improved |
| add test-time search with majority voting (NEO-S) | NEO · GridWorld alpha=0.33 Length OOD | transferability: 0.845 -> 0.907 | Improved |
| compositional multi-primitive programs instead of one monolithic latent | Disc-Mono vs NEO · GridWorld alpha=0.33 Comp. OOD | transferability: 0.000 -> 0.933 | Improved |
| test-time program optimization of continuous latent | Cont-Mono · GridWorld alpha=0.33 In-distribution | transferability: 0.001 -> 0.000 | No clear effect |
| increase alpha from 0.33 to 0.66 | NEO · Arithmetic Comp. OOD | transferability: 0.345 -> 0.573 | Improved |
Closest papers by meaning
- Eliciting Reasoning in Language Models with Cognitive Tools (NeurIPS 2025 · similarity 0.86)
- Does GPT Really Get It? A Hierarchical Scale to Quantify Human and AI’s Understanding of Algorithms (AAAI 2025 · similarity 0.84)
- Training Language Models to Explain Their Own Computations (NeurIPS 2026 · similarity 0.84)
- State of Thought Enables Endogenous Reasoning (NeurIPS 2026 · similarity 0.84)
- Hypothesis, Verification, and Induction: Grounding Large Language Models with Self-Driven Skill Learning (AAAI 2024 · similarity 0.84)
- Learning to Model the World With Language (ICML 2024 · similarity 0.84)
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.
