# EMBER: Efficient Memory via Budgeted Evidence Retention for Long-Horizon Agents

Source: [EMBER: Efficient Memory via Budgeted Evidence Retention for Long-Horizon Agents](<https://www.sota2.com/research/paper/6a22586e86173ad9c1468969>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-07-10T05:00:06.185Z

Authors: Yilong Li, Suman Banerjee, Tong Che

Published: 2026-06-04T09:03:05.000Z

## Abstract

Long-horizon agents can archive large histories, but future answers still incur retrieval, rereading, and context costs. When retained memory misses answer-relevant evidence, the system must return to larger portions of the raw history. We study budgeted evidence survival: before the query is known, which source evidence should be retained so that it remains recoverable and usable under a fixed retained source-evidence token budget? We instantiate this setting as Budgeted Pre-Query Retention, where memory is written during ingestion and later read without access to the full raw stream. We introduce EMBER, a learned retention policy that constructs a compact, source-backed evidence state. EMBER stores evidence capsules: verbatim source excerpts paired with retrieval keys and update metadata, preserving both grounding and read-time access. Post-query outcome feedback trains the writer to preserve evidence across the ingestion-retrieval-answer chain. On LongMemEval-RR, our LongMemEval-derived retained-evidence protocol, EMBER-14B reaches 0.3017 F1 at the 8192-token retained-evidence comparison point, compared with 0.1765 for the strongest non-EMBER budgeted baseline. Across retained source-evidence budgets, EMBER improves F1, Retain-Recall, and Read-Recall, indicating that long-horizon memory depends on retaining evidence within the budget rather than rereading larger histories.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Long-horizon Question Answering on LongMemEval-RR](<https://www.sota2.com/research/sota/long-horizon-question-answering-on-longmemeval-rr.md>): Long-horizon Question Answering; LongMemEval-RR; F1 Score: 30.17.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2606.05894>)

[PDF](<https://arxiv.org/pdf/2606.05894.pdf>)