Research map / Large language models / Language, safety and interpretability
CMedBench: A Comprehensive Benchmark for Efficient Medical Large Language Models
Shengbo Gao, Jinyang Guo, Lixian Su, Yifu Ding, Shiqiao Gu, Aishan Liu, Yuqing Ma, Zhiwang Zhang, Xianglong Liu
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 |
|---|---|---|---|
| apply 4-bit weight-only quantization (AWQ/GPTQ) | medical LLMs · medical knowledge tasks | CPS: CPS 0.99 (AWQ), 0.97 (GPTQ) | No clear effect |
| apply 4-bit SmoothQuant weight-activation quantization | medical LLMs · medical knowledge tasks | CPS: CPS 0.80 | Hurt |
| apply Wanda 50% unstructured sparsity | medical LLMs · medical knowledge tasks | CPS: CPS 0.90 | Hurt |
| apply ShortGPT 25% layer pruning | medical LLMs · medical knowledge tasks | CPS: CPS 0.74 | Hurt |
| apply Wanda 50% sparsity | medical LLMs · medical application tasks | CPS: CPS 0.96 | No clear effect |
| apply sparsity on Jmed dataset | medical LLMs · Jmed | accuracy | Hurt |
| scale up model size under ShortGPT | LLaMA3, Qwen2.5, HuatuoGPT-o1, Meditron · MMLU-Health | CCS: Qwen2.5-7B 0.87 -> Qwen2.5-72B 0.96 | Improved |
| apply 8-bit quantization instead of 4-bit | GPTQ/AWQ/SmoothQuant · MMLU-Health | average normalized ratio: SmoothQuant 0.77 -> 1.00 | Improved |
| apply Wanda sparsity 25% to 75% | medical LLMs · MMLU-Health | average normalized ratio: 0.99 -> 0.69 | Hurt |
| apply quantization | medical LLMs · trustworthiness (Fairness) | TWY | Mixed |
| apply unstructured sparsity Wanda | vLLM inference · efficiency | speed and memory | No clear effect |
| apply quantization | medical LLMs · efficiency | memory and token rate | Improved |
Closest papers by meaning
- LLMCBench: Benchmarking Large Language Model Compression for Efficient Deployment (NeurIPS 2024 · similarity 0.90)
- MedJourney: Benchmark and Evaluation of Large Language Models over Patient Clinical Journey (NeurIPS 2024 · similarity 0.89)
- Medmarks: A Comprehensive Open-Source LLM Benchmark Suite for Medical Tasks (NeurIPS 2026 · similarity 0.89)
- MedBench: A Large-Scale Chinese Benchmark for Evaluating Medical Large Language Models (AAAI 2024 · similarity 0.88)
- MedCalc-Bench: Evaluating Large Language Models for Medical Calculations (NeurIPS 2024 · similarity 0.87)
- Compressing LLMs: The Truth is Rarely Pure and Never Simple (ICLR 2024 · similarity 0.87)
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.
