Research map / Vision-language and multimodal / Multimodal LLMs
Circuit-MLLM: Topological Logic-Guided Latent-Space Visual Reasoning for Circuit Schematic Understanding
Jinyuan Deng, Yuqi Jiang, Wenjing Huang, Xin Li, Qi Sun, Cheng Zhuo
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 topology-guided sequencing (ablation) | Latent reasoning 7B · Circuit-MLLM-bench connection tasks (Easy/Medium/Hard) | Avg: 69.85 -> 70.61 | Improved |
| add circuit-expert distillation (ablation) | Latent reasoning 7B · Circuit-MLLM-bench connection tasks (Easy/Medium/Hard) | Avg: 69.85 -> 70.64 | Improved |
| combine sequencing and expert distillation (ablation) | Latent reasoning 7B · Circuit-MLLM-bench connection tasks (Easy/Medium/Hard) | Avg: 70.61 -> 72.26 | Improved |
| use latent size 4 instead of 2 (ablation) | Circuit-MLLM · Circuit-MLLM-bench | Avg score: 74.42 -> 76.20 | Improved |
| use latent size 4 instead of 6 (ablation) | Circuit-MLLM · Circuit-MLLM-bench | Avg score: 74.19 -> 76.20 | Improved |
| use latent size 4 instead of 8 (ablation) | Circuit-MLLM · Circuit-MLLM-bench | Avg score: 75.43 -> 76.20 | Improved |
| add HAWP and DeepLSD to DINOv2 experts (ablation) | Circuit-MLLM · Circuit-MLLM-bench | Avg score: 75.16 -> 76.20 | Improved |
| add SFT to Qwen2.5-VL 7B | Qwen2.5-VL 7B · Circuit-MLLM-bench connection tasks (Easy/Medium/Hard) | Avg: 45.09 -> 67.74 | Improved |
| add GRPO after SFT | Qwen2.5-VL 7B · Circuit-MLLM-bench connection tasks (Easy/Medium/Hard) | Avg: 67.74 -> 67.75 | No clear effect |
| add Visual CoT cropping | GPT-4o · Connection tasks, CoT comparison | Avg: 59.57 -> 56.74 | Hurt |
| replace textual CoT with latent-visual reasoning (ablation) | Qwen2.5-VL 7B · Connection tasks, CoT comparison | Avg: 70.75 -> 72.26 | Improved |
| use latent reasoning vs Mirage | Circuit-MLLM · Circuit-MLLM-bench connection tasks (Easy/Medium/Hard) | Avg: 69.82 -> 72.26 | Improved |
Closest papers by meaning
- CircuitSense: A Hierarchical Circuit System Benchmark Bridging Visual Comprehension and Symbolic Reasoning in Engineering Design Process (ICLR 2026 · similarity 0.90)
- AnalogToBi: Device-Level Analog Circuit Topology Generation via Bipartite Graph and Grammar Guided Decoding (NeurIPS 2026 · similarity 0.87)
- CircuitFusion: Multimodal Circuit Representation Learning for Agile Chip Design (ICLR 2025 · similarity 0.87)
- Efficient Automated Circuit Discovery in Transformers using Contextual Decomposition (ICLR 2025 · similarity 0.86)
- AutoVSR: Automatic Visual-to-Symbolic Reasoning for Symbolic Expression Generation from Circuit Schematic (ICML 2026 · similarity 0.85)
- DeepGate4: Efficient and Effective Representation Learning for Circuit Design at Scale (ICLR 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.
