# ConvMemory v2: A Recall-Preserving Top-10 Evidence Reranker for Conversational Memory Retrieval

arXiv: https://arxiv.org/abs/2606.10842 · HTML: https://mlsanity.com/paper/2606.10842/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Builds on 3 of its 3 references that are in the map.
- First posted on arXiv 9 Jun 2026.

## Builds on (most cited first, 3 of 3)

- [MPNet: Masked and Permuted Pre-training for Language Understanding](https://mlsanity.com/paper/2004.09297/) (2020)
- [Evaluating Very Long-Term Conversational Memory of LLM Agents](https://mlsanity.com/paper/2402.17753/) (2024)
- [ConvMemory: A Lightweight Learned Memory Reranker, a Negative Attribution Result, and a Research-Preview Conflict Editor](https://mlsanity.com/paper/2605.28062/) (2026)

## Built on by (most cited first, 2 of 2)

- [When Does Selection Replace Extraction? A Pre-Registered Test of Agent Memory with a Typed Decision Model](https://mlsanity.com/paper/2609.34227/) (2026)
- [ConvMemory v3: A Validity Context Layer for Conversational Memory via Target-Conditioned Relation Verification](https://mlsanity.com/paper/2606.26753/) (2026)

## Abstract (arXiv)

We describe ConvMemory v2, an opt-in token-evidence reranker that sits after the lightweight ConvMemory v1 reranker and reorders only v1's protected top-10 candidate set. v2 is a fine-tuned ms-marco-MiniLM-L-6-v2 cross-encoder (22,713,601 parameters, measured from the released checkpoint) applied to the ten (query, memory) pairs that v1 has already selected; it does not change which ten memories are returned, so Recall@10 and Hit@10 are identical to v1 by construction, not by statistical coincidence. On the LoCoMo conversational memory benchmark (5 seeds, n = 4955 test rows), v2 raises FULL MRR from v1's 0.5824 to 0.6560 (paired bootstrap +0.0734, 95% CI [+0.0645, +0.0827]) and H@1 from 0.4440 to 0.5474. v2 closes most but not all of the gap to a much more expensive full-pool cross-encoder reference (mxbai-rerank-large-v1 over the top-500, MRR 0.6688): on FULL MRR v2 sits 0.013 below mxbai_top500, but on two raw-dense-hard slices (where v1's protected top-10 has higher recall than mxbai's own top-10) v2 exceeds mxbai_top500. A four-arm load-bearing ablation shows candidate-specific memory text is the mechanism: removing, shuffling, or replacing it collapses MRR below raw dense retrieval. v2 is best understood as a standard recall-preserving cascade pattern with LoCoMo-specific fine-tuning, an explicit anti-shortcut inference contract, and disciplined load-bearing analysis; its advantage over mxbai is slice-specific rather than a general dominance claim. This report extends the v1 technical report (arXiv:2605.28062).