LEMMA: Learned Guidance for Evidence-Carrying Long-Horizon Symbolic Rewriting

LEMMA is a neuro-symbolic system for long-horizon symbolic mathematical rewriting. It combines a symbolic rule engine, guardrail-filtered action generation, evidence-typed verification, and learned policy guidance to transform a starting expression into a specified goal through individually validated rewrite steps.The accompanying study evaluates six search and decoding procedures on a frozen 120-problem benchmark with reference derivations of depths 64, 96, and 128. On the 89 problems containing genuine branching decisions, learned local action ranking substantially improves fixed-budget search efficiency.

Authors

Institutions

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-09-19
DOI
https://doi.org/10.5281/zenodo.22842820
Primary Topic
Mathematics, Computing, and Information Processing
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
article

LEMMA: Learned Guidance for Evidence-Carrying Long-Horizon Symbolic Rewriting

Pushp Kharat, Atul Saxena
Zenodo (CERN European Organization for Nuclear Research)
Mathematics, Computing, and Information Processing
article

LEMMA: Learned Guidance for Evidence-Carrying Long-Horizon Symbolic Rewriting

Pushp Kharat, Atul Saxena
article en

Abstract

LEMMA is a neuro-symbolic system for long-horizon symbolic mathematical rewriting. It combines a symbolic rule engine, guardrail-filtered action generation, evidence-typed verification, and learned policy guidance to transform a starting expression into a specified goal through individually validated rewrite steps.The accompanying study evaluates six search and decoding procedures on a frozen 120-problem benchmark with reference derivations of depths 64, 96, and 128. On the 89 problems containing genuine branching decisions, learned local action ranking substantially improves fixed-budget search efficiency.

Zenodo (CERN European Organization for Nuclear Research)
Kubota (Japan) (JP), SciGenom Labs (India) (IN)
Openalex Percentile: Top 9%
Mathematics, Computing, and Information Processing
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.