DecisionRAG paper: decision models replace retrievers, beating 8B rerankers by 13.8 R@1
_reachsumit · x · 2026-10-28
Researchers including Kilian Weinberger's group at Cornell propose DecisionRAG, replacing the retriever with a "decision model": like a person browsing a library, it navigates the corpus's inherent hierarchy from broad topics down to passages, scoring all candidate branches at each level in a single forward pass — generating no tokens.
Key results:
- Without task-specific training or a first-stage retriever, it beats dense retrieval with an 8B reranker by 13.8 R@1 on MultiDoc2Dial and 17.4 Hit@1 on Natural Questions
- Across four benchmarks including a 2.68M-paragraph Wikipedia corpus, retrieval takes just 0.7–1.8 seconds per query
The motivation is fixing the retrieval-generation capability asymmetry: lexical/embedding-based retrievers miss what strong generators could otherwise use; judging candidates in shared context makes relevance estimates comparable.
More from Research
- Ex-OpenAI researcher's 26,000-word treatise claims AI's hardest interpretability problem just got easier — nordicinst · 2026-10-28
- Researchers Build an Adaptive AI Interviewer to Scale Occupational Task Inventories — soumitrashukla9 · 2026-10-28
- Snowflake Doubles Down on Agentic RL: Overfit First, Then Scale to 65.7 on SWE-bench — StasBekman · 2026-10-28
- Force field experts debate: zinc coordination fitting and non-scalable empirical parameterization — TimothyDuignan · 2026-10-28
- TTME beats test-time training on 5 math tasks while cutting FLOPs by 82.8-84.6% — UCSantaBarbara · 2026-10-28
- EnvLunch turns executable environments into free verifiable agent training data — KingsCollegeLondon · 2026-10-28