Research map / Large language models / Agents, code and math
Symbol-LLM: Leverage Language Models for Symbolic System in Visual Human Activity Reasoning
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 Symbol-LLM reasoning to CLIP | CLIP · HICO | mAP: 67.12 -> 69.73; 37.08 -> 43.21 | Helped |
| Add Symbol-LLM reasoning to CLIP | CLIP · HAKE-Verb | mAP: 73.82 -> 75.27; 43.92 -> 48.95 | Helped |
| Add Symbol-LLM reasoning to CLIP | CLIP · Stanford-40 | mAP: 75.68 -> 82.22 | Helped |
| Add Symbol-LLM reasoning to CLIP | CLIP · HAKE-PaSta | accuracy: 39.36 -> 40.47 | Helped |
| Add Symbol-LLM reasoning to BLIP2 | BLIP2 · HICO | mAP: 50.61 -> 53.15 | Helped |
| Add Symbol-LLM reasoning to BLIP2 | BLIP2 · HAKE-Verb | mAP: 49.47 -> 51.37 | Helped |
| Add Symbol-LLM reasoning to BLIP2 | BLIP2 · Stanford-40 | mAP: 91.85 -> 92.59 | Helped |
| Add Symbol-LLM reasoning to BLIP2 | BLIP2 · HAKE-PaSta | accuracy: 43.81 -> 44.65 | Helped |
| Use symbolic reasoning instead of trivially combining CLIP and BLIP2 predictions | CLIP+Reasoning · HAKE-Verb | mAP: 44.76 -> 48.95 | Helped |
| Use the proposed broad-coverage, rational symbolic system instead of HAKE rules (ablation) | CLIP+Reasoning · HAKE-Verb | mAP: 44.98 -> 48.95 | Helped |
| Add logical entailment checking to rules (ablation) | Symbol-LLM · HAKE-Verb | mAP: 46.51 -> 48.95 | Helped |
| Add the symbol-rule loop to diversify generated rules (ablation) | Symbol-LLM · HAKE-Verb | mAP: 47.53 -> 48.95 | Helped |
Closest papers by meaning
- Can Large Language Models Understand Symbolic Graphics Programs? (ICLR 2025 · similarity 0.87)
- Visual symbolic mechanisms: Emergent symbol processing in Vision Language Models (ICLR 2026 · similarity 0.86)
- Concept-RuleNet: Grounded Multi-Agent Neurosymbolic Reasoning in Vision Language Models (AAAI 2026 · similarity 0.85)
- MentisOculi: Revealing the Limits of Reasoning with Mental Imagery (ICML 2026 · similarity 0.85)
- CodeDance: A Dynamic Tool-integrated MLLM for Executable Visual Reasoning (CVPR 2026 · similarity 0.85)
- Visual Graph Arena: Evaluating Visual Conceptualization of Vision and Multimodal Large Language Models (ICML 2025 · similarity 0.85)
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.