AI Paper Atlas

Research map / Large language models / Retrieval and knowledge

SmartChunk Retrieval: Query-Aware Chunk Compression with Planning for Efficient Document RAG

Xuechen Zhang, Koustava Goswami, Samet Oymak, Jiasi Chen, Nedim Lipka

ICLR 2026 · Large language models › Retrieval and knowledge · approach: RAG robustness and evaluation · 1 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add the query-aware planner (ablation)SmartChunk · Main QA benchmarksmonetary cost / latency: cost 0.096 -> 0.078, latency 1.94 -> 3.62 for full modelImproved
use trained compression encoder instead of direct encoding of chunks (ablation)SmartChunk · Main QA benchmarksQA accuracy: 0.427 -> 0.564Improved
use compression encoder instead of GPT summarization of each chunk (ablation)SmartChunk · Main QA benchmarksmonetary cost: 0.204 -> 0.078Improved
use SmartChunk instead of single-level fixed-size chunkingRAG pipeline · Main QA benchmarksQA accuracy: 0.285 -> 0.564Improved
use few-shot prompting for planner on out-of-domain dataSmartChunk · NewsQAF1: 0.875 -> 0.906Improved
fine-tune planner with reasoning instead of frozen LLM (ablation)planner · Planner evaluationplanning accuracy: 0.426 -> 0.820Improved
add reasoning to fine-tuned planner (vs SLM classifier) (ablation)planner · Planner evaluationplanning accuracy: 0.724 -> 0.820Improved
train planner with STITCH instead of SFT+RL (ablation)planner · Planner training (418k tokens)planning accuracy: 0.544 -> 0.820Improved
use RL only without SFT or hints (ablation)planner · Planner trainingplanning accuracy: 0.820 -> 0.356Hurt
sample reasoning traces from 6 diverse LLMs instead of one (ablation)planner SFT · Planner trace diversityplanning accuracy: 53.1% -> 74.3%Improved
use a larger single trace-generating model (ablation)planner SFT · Planner trace diversityplanning accuracy: 53.1% -> 45.6%Hurt
combine SmartChunk with late chunking and hybrid searchSmartChunk · QA benchmarksQA accuracyImproved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy