Paper

When Retrieval Doesn't Help: A Large-Scale Study of Biomedical RAG

Unreadunread

AI summary

arXiv:2606. 04127v1 Announce Type: new Abstract: Medical question answering is a high-stakes setting where factual errors can have serious consequences.

Reading status

Priority

About this paper

Authors: Erfan Nourbakhsh, Rocky Slavin, Ke Yang, Anthony Rios
Released: 2026-06-04 (2mo ago)
Topic: Related to current project
Relation to my work: Read next because When Retrieval Doesn't Help: A Large-Scale Study of Biomedical RAG overlaps with clean result "Only continuous soft prefixes hit both EM axes at once on Qwen-2.5-7B-Instruct: discrete prompt searches split between the alignment objective and the distributional objective, and both discretizations of the soft prefix collapse (MODERATE confidence)", clean result "Training one persona to emit a [ZLT] marker without bystanders adopting it has a one-cell-wide LR x epochs window on Qwen2.5-7B-Instruct (LOW confidence)", clean result "The marker is a representational handle, not a behavioural one — sharing it between a villain persona and the assistant transfers no misalignment (HIGH confidence)". Matching terms: eval, source, line, alone, does, model. Source: arxiv cs.CL (NLP).
arXiv: 2606.04127

Ask Claude about this paper

Posts a comment with @claude so the runner spins up a Claude Code session scoped to When Retrieval Doesn't Help: A Large-Scale Study of Biomedical RAG. Replies appear in the conversation below.

Comments

No comments yet.

Related records

No links yet.