Research map / Large language models / Agents, code and math
Knowledge-Centric Agents for Workflow Generation in ComfyUI
Zhendong Li, Lei Sun, Ruibo Ming, He Zhang, Danda Paudel, Luc Van Gool, Jinjin Gu
This paper tried a common change: Chain-of-thought prompting
add CoT or self-consistency prompting to GPT-5 · Own test set (30 workflows) · TaskCons (%): 34.7 (Few-Shot) vs 32.9 (CoT) vs 34.8 (CoT-SC) No clear effect
What 1,107 accepted papers report for this change
- 765 improved
- 94 no clear effect
- 52 mixed
- 196 hurt
- AudioChat: Unified Audio Storytelling, Editing, and Understanding with Transfusion Forcing ICML 2026 · add structured chain-of-thought reasoning to the diffusion LLM · StoryGen-Eval editing · delta multiFLAM: 27.8 -> 11.7Improved
- Vision Language Models Cannot Reason About Physical Transformation ICML 2026 · add CoT prompt instead of Direct prompt · Number & Length · accuracy: F(3, 333) = 18.28Hurt
- VideoSEG-O3: A Multi-turn Reinforcement Learning Framework for Reasoning Video Object Segmentation ICML 2026 · Add CoT cold-start stage on top of SFT · Long-RVOS · J&F: 47.12 -> 51.82Improved
See this change in detailMatched by the wording of each paper's own result, read by AI from the full text. Papers report gains more often than failures, so this shows what gets published, not the chance a change works.
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 |
|---|---|---|---|
| use hierarchical strategy/skeleton supervision with Qwen3-14B instead of end-to-end Qwen3-32B Q2P (ablation) | knowledge-centric agent · Own test set (30 workflows) | Pass (%): 56.7 -> 86.9 | Improved |
| use hierarchical strategy supervision instead of end-to-end Q2P (ablation) | knowledge-centric agent · Own test set (30 workflows) | VND: 148 -> 152 | Improved |
| use hierarchical strategy supervision instead of end-to-end Q2P (ablation) | knowledge-centric agent · Own test set (30 workflows) | NodeComp (%): 92.3 -> 96.3 | Improved |
| use knowledge-centric agent instead of ComfyAgent | knowledge-centric agent · Own test set (30 workflows) | Resolve (%): 29.4 -> 61.1 | Improved |
| use knowledge-centric agent instead of RAG | knowledge-centric agent · Own test set (30 workflows) | VND: 42 -> 152 | Improved |
| evaluate on ComfyBench subset | knowledge-centric agent · ComfyBench subset | Resolve (%): 25.0 (ours) vs 25.0 (RAG) | No clear effect |
| evaluate on FlowBench | knowledge-centric agent · FlowBench | Pass (%): 46.7 -> 66.7 | Improved |
| swap LLM judge from GPT-5 to Gemini-2.5 (ablation) | TaskCons evaluation · Own test set (30 workflows) | TaskCons (%): 86.9 vs 85.6 | No clear effect |
| add CoT or self-consistency prompting to GPT-5 (ablation) | GPT-5 baselines · Own test set (30 workflows) | TaskCons (%): 34.7 (Few-Shot) vs 32.9 (CoT) vs 34.8 (CoT-SC) | No clear effect |
| add retrieval of exemplars (RAG) to GPT-5 (ablation) | GPT-5 baselines · Own test set (30 workflows) | NodeComp (%): 36.7 -> 56.7 | Improved |
Closest papers by meaning
- ComfyMind: Toward General-Purpose Generation via Tree-Based Planning and Reactive Feedback (NeurIPS 2025 · similarity 0.87)
- Benchmarking Agentic Workflow Generation (ICLR 2025 · similarity 0.87)
- GraphFlow: A Graph-Based Workflow Management for Efficient LLM-Agent Serving (ICML 2026 · similarity 0.86)
- WorkArena++: Towards Compositional Planning and Reasoning-based Common Knowledge Work Tasks (NeurIPS 2024 · similarity 0.86)
- FlowSteer: Towards Agents Designing Agentic Workflows via Reinforced Progressive Canvas Editing (NeurIPS 2026 · similarity 0.86)
- DyFlow: Dynamic Workflow Framework for Agentic Reasoning (NeurIPS 2025 · 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.
