AI Paper Atlas

Research map / Learning theory and algorithms / Combinatorial and search optimization

SoftMatcha 2: A Fast and Soft Pattern Matcher for Trillion-Scale Corpora

Masataka Yoneda, Yusuke Matsushita, Go Kamoda, Kohei Suenaga, Takuya Akiba, Masaki Waga, Sho Yokoi

ICML 2026 · Learning theory and algorithms › Combinatorial and search optimization · approach: Heuristic and local search algorithms · 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
use disk-aware staged suffix array instead of standard suffix array (infini-gram)exact lookup · FineWeb-Edu 1.4Tp95 latency (ms): 11.05 -> 0.34Improved
use disk-aware staged suffix array instead of FM-index (infini-gram mini)exact lookup · FineWeb-Edu 273Bp95 latency (ms): 151.84 -> 0.32Improved
add all three pruning techniques (iterative, k-gram, last-bits) (ablation)SoftMatcha 2 soft search · FineWeb-Edugeometric mean lookup growth per token: 5.17 -> 2.30Improved
add run-length compression to the index (ablation)SoftMatcha 2 index · FineWeb-Edu 1.4Tindex size (TB): 56.0 -> 21.6Improved
replace inverted index with suffix array and pruningSoftMatcha · Soft search latency vs corpus sizelatency: SoftMatcha latency increases almost linearly beyond 1B tokens; ours shows almost no increaseImproved
use SoftMatcha 2 instead of SoftMatchasoft search · LLM-as-a-judge qualitywin rate (%): 61.3Improved
use soft search in addition to exact matchcontamination detection · 7 benchmarkssamples flagged dirty: 338 of 2,564 (13.2%) -> additional 36 (1.4%), 29/36 true (81%)Improved
use SoftMatcha 2 instead of SoftMatchainformation retrieval · TREC-COVIDP@20: 35.4 -> 36.0Improved
use SoftMatcha 2 instead of BM25information retrieval · SciFactP@1: 42.7 -> 42.7No clear effect
use SoftMatcha 2 instead of BM25paraphrase detection · Paraphrase detectionscore: 82.8 -> 84.3Improved
use SoftMatcha 2 instead of infini-gram exact matchparaphrase detection · Paraphrase detectionscore: 65.5 -> 84.3Improved

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