AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Self-Improving Sparse Retrieval Through Heuristic Representation Refinement and Representation-Focused Learning

Xiaojing Li, Bin Wang, Xiaochun Yang, Meng Luo

AAAI 2026 · Architectures and efficiency › RNNs and normalization · approach: Learned embeddings for retrieval · 0 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 irrelevant-token removal (ablation)SIRE (SPLADE++) · MS MARCO devMRR@10: 40.7 -> 42.1Improved
add irrelevant-token removal (ablation)SIRE (SPLADE++) · BEIRNDCG@10: 49.0 -> 51.3Improved
add low-impact token reduction (ablation)SIRE (SPLADE++) · BEIRNDCG@10: 50.2 -> 51.3Improved
add relevant-token augmentation with LLM docs (ablation)SIRE (SPLADE++) · BEIRNDCG@10: 49.4 -> 51.3Improved
add query KL representation loss (ablation)SIRE (SPLADE++) · MS MARCO devMRR@10: 40.2 -> 42.1Improved
add document contrastive loss (ablation)SIRE (SPLADE++) · MS MARCO devMRR@10: 40.6 -> 42.1Improved
keep retrieval loss alongside representation losses (ablation)SIRE (SPLADE++) · MS MARCO devMRR@10: 38.8 -> 42.1Improved
iterate refinement and learning up to 3 rounds (ablation)SIRE (SPLADE++) · BEIRNDCG@10: 49.4 -> 51.3Improved
iterate refinement and learning (ablation)SIRE (SPLADE++) · MS MARCO devMRR@10: 40.9 -> 42.1Improved
apply heuristic refinement to SPLADE queries at inferenceSPLADE · BEIRNDCG@10: 43.2 -> 44.9Improved
apply heuristic refinement to SPLADE-v3 queriesSPLADE-v3 · BEIRNDCG@10: 48.2 -> 50.9Improved
apply heuristic refinement to SPLADE++ queriesSPLADE++ · MS MARCO devMRR@10: 38.0 -> 39.3Improved

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