AI Paper Atlas

Research map / Large language models / Retrieval and knowledge

Memora: A Harmonic Memory Representation Balancing Abstraction and Specificity

Menglin Xia, Xuchao Zhang, Shantanu Dixit, Paramaguru Harimurugan, Rujia Wang, Victor Rühle, Robert Sim, Chetan Bansal, Saravan Rajmohan

ICML 2026 · Large language models › Retrieval and knowledge · approach: Retrieval-augmented reasoning · 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
add primary abstraction layer (ablation)Mem0-style memory · LoCoMo overallLLM-as-a-Judge: 0.653 -> 0.795Improved
add memory update (merging) (ablation)Memora with primary abstraction · LoCoMo overallLLM-as-a-Judge: 0.795 -> 0.801Improved
add cue anchors, semantic retriever (ablation)Memora (S) · LoCoMo overallLLM-as-a-Judge: 0.850 -> 0.849No clear effect
add cue anchors, policy retriever (ablation)Memora (P) · LoCoMo overallLLM-as-a-Judge: 0.851 -> 0.863Improved
use policy instead of semantic retrieverMemora · LoCoMo overallLLM-as-a-Judge: 0.849 -> 0.863Improved
use raw segments instead of extracted episodic memory (ablation)Memora (P) · LoCoMo overallLLM-as-a-Judge: 0.838 -> 0.863Improved
add factual memory to episodic-only (ablation)Memora (P) · LoCoMo overallLLM-as-a-Judge: 0.851 -> 0.863Improved
add episodic memory to factual only (ablation)Memora (P) · LoCoMo overallLLM-as-a-Judge: 0.833 -> 0.863Improved
use policy retriever instead of semanticMemora · LoCoMo latencyend-to-end mean latency: 1.062s -> 5.697sHurt
predict index offsets for constructionMemora · LoCoMo construction timeseconds: 1322.0 -> 739.9Improved
replace construction LLM gpt-4.1-mini with gpt-5.4-nano (ablation)Memora (S) · LoCoMo overallLLM-as-a-Judge: 0.849 -> 0.763Hurt
replace construction LLM gpt-4.1-mini with gpt-5.4-nano (ablation)Memora (P) · LoCoMo overallLLM-as-a-Judge: 0.863 -> 0.851No clear effect

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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