PulseAugur
EN
LIVE 13:56:08

ConvMemory: Lightweight AI Memory Reranker Offers Cost-Effective Performance

A new research paper introduces ConvMemory, a lightweight 3.6M-parameter reranker designed for conversational long-term memory retrieval. The model demonstrates competitive performance against larger cross-encoders in terms of recall and latency, while operating at a significantly lower cost. The research also includes a negative attribution result, clarifying that ConvMemory's mechanism is primarily distillation rather than temporal-structure exploitation, and releases CCGE-LA, a conflict-aware candidate-set editor. AI

IMPACT This research offers a more cost-efficient approach to long-term memory retrieval in conversational AI systems.

RANK_REASON The cluster contains a research paper detailing a new AI model and its performance evaluation.

Read on arXiv cs.IR (Information Retrieval) →

AI-generated summary · Google Gemini · from 2 sources. How we write summaries →

ConvMemory: Lightweight AI Memory Reranker Offers Cost-Effective Performance

COVERAGE [2]

  1. arXiv cs.CL TIER_1 English(EN) · Taiheng Pan ·

    ConvMemory: A Lightweight Learned Memory Reranker, a Negative Attribution Result, and a Research-Preview Conflict Editor

    arXiv:2605.28062v1 Announce Type: new Abstract: We describe ConvMemory, a small 3.6M-parameter learned reranker for conversational long-term memory retrieval, trained with cross-encoder teacher supervision over fused dense and lexical features. On the LongMemEval memory family, C…

  2. arXiv cs.IR (Information Retrieval) TIER_1 English(EN) · Taiheng Pan ·

    ConvMemory: A Lightweight Learned Memory Reranker, a Negative Attribution Result, and a Research-Preview Conflict Editor

    We describe ConvMemory, a small 3.6M-parameter learned reranker for conversational long-term memory retrieval, trained with cross-encoder teacher supervision over fused dense and lexical features. On the LongMemEval memory family, ConvMemory operates above the BGE-large cross-enc…