Medical retrieval model trained on RTX 3090 rivals full dataset runs
tomaarsen · x · 2026-08-26
Tomaarsen released the mLateOn-medical model and a detailed blog post. Scaling experiments revealed that training 100k pairs for 75 minutes on an RTX 3090 achieves within 1.2 NDCG@10 points of the full million-pair run. The model utilizes Multi-Vector Encoder and shows strong performance on medical retrieval benchmarks.
More from Apps
- Build LEGO in the Browser with LLM, LDraw, and Three.js — nptacek · 2026-08-26
- Basecamp goes Agent-Native with CLI and Agent Skill for AI automation — danshipper · 2026-08-26
- Each::labs MCP: Control Premiere Pro via text prompts — aziz4ai · 2026-08-26
- Inth scans your codebase, site and PRs to catch privacy risks before they block deals — paw_lean · 2026-08-26
- OpenBrief: Transcribe video/audio locally and chat with grounded summaries — tom_doerr · 2026-08-26
- Do users struggle to explain intent to AI? Discussion on prompt barriers. — SidequestMatt · 2026-08-26