Research map / Large language models / Agents, code and math
Surgical AI Copilot: Energy-Based Fourier Gradient Low-Rank Adaptation for Surgical LLM Agent Reasoning and Planning
Jiayuan Huang, Runlong He, Danyal Zaman Khan, Evangelos B. Mazomenos, Danail Stoyanov, Hani Marcus, Linzhe Jiang, Matthew John Clarkson, Mobarak I. Hoque
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 |
|---|---|---|---|
| use DEFT-GaLore instead of GaLore | LLaMA-3.2-3B planner · PitAgent | BLEU-4 (%): 70.68 -> 71.88 | Improved |
| use DEFT-GaLore instead of GaLore | LLaMA-3.2-3B planner · PitAgent | model selection ACC (%): 94.94 -> 96.25 | Improved |
| use DEFT-GaLore instead of GaLore | Qwen-2.5-1.5B planner · PitAgent | BLEU-4 (%): 47.68 -> 55.53 (text says +8.85) | Mixed |
| use DEFT-GaLore instead of GaLore | Qwen-2.5-1.5B planner · PitAgent | F1 (%): 88.67 -> 95.33 (text says +12.89) | Mixed |
| use DEFT-GaLore instead of GaLore (training time) | LLaMA-3.2-3B · PitAgent | training time per epoch (s): 8762.43 -> 3355.75 | Improved |
| use DEFT-GaLore instead of GaLore (training time) | Qwen-2.5-1.5B · PitAgent | training time per epoch (s): 2733.62 -> 1764.88 | Improved |
| use DEFT-GaLore instead of DCT-GaLore | LLaMA-3.2-3B planner · PitAgent | training time per epoch (s): 3535.69 -> 3355.75 | Improved |
| use DEFT-GaLore instead of GaLore 2 | LLaMA-3.2-3B planner · PitAgent | model selection ACC (%): 94.85 -> 96.25 | Improved |
| use GaLore-family projection instead of LoRA/DoRA/MoRA | LLaMA-3.2-3B planner · PitAgent | BLEU-4 (%): 46.32 (DoRA) -> 70.68 (GaLore) | Improved |
| use DEFT-GaLore rank 64 instead of 128 (ablation) | DEFT-GaLore · PitAgent | BLEU-4 (%): 71.88 -> 65.71 | Hurt |
| use DEFT-GaLore rank 256 instead of 128 (ablation) | DEFT-GaLore · PitAgent | BLEU-4 (%): 71.88 -> 69.04 | Hurt |
| use DEFT-GaLore prompts instead of GaLore prompts for downstream VQA | PitVQA++ Worker · PitVQA | BLEU-4 (%): 67.31 -> 70.73 | Improved |
Closest papers by meaning
- Simple Agents Outperform Experts in Biomedical Imaging Workflow Optimization (CVPR 2026 · similarity 0.83)
- SurgCoT: Advancing Spatiotemporal Reasoning in Surgical Videos through a Chain-of-Thought Benchmark (CVPR 2026 · similarity 0.83)
- MDAgents: An Adaptive Collaboration of LLMs for Medical Decision-Making (NeurIPS 2024 · similarity 0.82)
- SurgicAI: A Hierarchical Platform for Fine-Grained Surgical Policy Learning and Benchmarking (NeurIPS 2024 · similarity 0.82)
- Multi-scale Temporal Prediction via Incremental Generation and Multi-agent Collaboration (NeurIPS 2025 · similarity 0.82)
- SurgVista: Long-Horizon Surgical World Modeling with Plausible Instrument-Tissue Dynamics (NeurIPS 2026 · similarity 0.82)
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.
