Claude Code · Open Source · Claude · AMD · Hugging Face
"OncoAgent: A Dual-Tier Multi-Agent Framework for Privacy-Preserving Oncology Clinical Decision Support"
Compiled by KHAO Editorial — aggregated from 1 outlet. See llms.txt for citation guidance.
★ Tier-1 Source
Technical preprint · May 2026 · OncoAgent Research Group.
Key facts
- Decision boundary: S ≥ 0.5 → Tier 2 (Qwen 3.6-27B deep reasoning)
- S < 0.5 → Tier 1 (Qwen 3.5-9B speed triage)
- After migrating from Qwen 3.5 to Qwen 2.5 Instruct for the grading step, success rate improved from 0% → 100%, with RAG confidence score reaching 2.3+ on uterine cancer triage tests
- The knowledge base was constructed from 77 direct physician guideline PDFs identified by a concurrent web scraper that processed 138 NCCN detail pages in under 60 seconds
- Checkpoint-1000 results: Tier 1 adapter trained for 1,339 steps
- training loss ≈ 0.05
- adapter size 187 MB
- verified against 11-file manifest including adapter_model.safetensors, adapter_config
Summary
The team present OncoAgent, an open-source, privacy-preserving clinical decision support system for oncology. The system routes clinical queries through an additive complexity scorer to either a 9B parameter speed-optimised model (Tier 1) or a 27B deep-reasoning model (Tier 2), both fine-tuned via QLoRA on a corpus of 266,854 real and synthetically generated oncological cases using the Unsloth framework on AMD Instinct MI300X hardware (192 GB HBM3). Sequence packing on MI300X enabled full-dataset fine-tuning in approximately 50 minutes enabling hospital deployment without data exfiltration. Large language models have demonstrated significant promise in clinical NLP tasks including diagnostic coding, literature summarisation, and patient communication. Decomposed multi-agent systems have emerged as a principled approach to complex reasoning tasks.