Alibaba's QESChunker unifies overlapping and non-overlapping RAG chunking via question-evidence supervision
_reachsumit · x · 2026-10-29
Alibaba researchers propose QESChunker, reportedly the first method to build static chunk indices from question-evidence supervision without access to test queries.
- Recasts RAG chunking as a supervised combinatorial optimization problem maximizing question-evidence utility, using human-annotated or LLM-generated question-evidence pairs.
- Under the non-overlap constraint, the objective decomposes per chunk and is solved exactly in polynomial time via dynamic programming.
- With overlap allowed, it becomes monotone submodular (exact optimization NP-hard); their greedy achieves a (1-1/√e)-approximation.
- Across two self-built and three public datasets, both variants beat existing chunkers in complete-evidence recall in nearly all settings, under the same overlap regime and a fixed hybrid retriever.
16 pages, 8 figures, 4 tables.
More from Research
- Granary-v2 open speech model expands to 48 languages with open data and pipeline — rdesh26 · 2026-10-29
- Tinker launches Dimuon, a Muon-based optimizer that beats AdamW on LoRA fine-tuning — xennygrimmato_ · 2026-10-29
- Six CUDA kernel optimizations squeeze another 1.78× speedup out of 3DGS training — gkopanas · 2026-10-29
- Cross-lingual LLM effects paper wins Outstanding Paper award at EMNLP 2026 — anas_ant · 2026-10-29
- Lie detector compiled into Qwen3.5 weights flags 91% of lies with zero false positives — teortaxesTex · 2026-10-29
- Practitioner: No serious lab actually trains on the test set today — nrehiew_ · 2026-10-29