Capability-Driven Self-Evolution of Agent Memory

Yaoqi Chen; Yuru Feng; Qianxi Zhang; Baotong Lu; Jianan Lu; Zhirui Wang; Shusen Xu; Zewen Jin

cs.AIOct 5, 2026

Capability-Driven Self-Evolution of Agent Memory

Authors: Yaoqi Chen, Yuru Feng, Qianxi Zhang, Baotong Lu, Jianan Lu, Zhirui Wang, Shusen Xu, Zewen Jin, +3 more

Organizations: University of Science and Technology of China · Microsoft · University of California, San Diego

Abstract

Memory self-evolution uses task feedback to iteratively improve executable memory programs that store and retrieve information from past interactions. Existing approaches typically adopt holistic evolution, deriving revision directions from mixed feedback and judging progress by overall performance. This can obscure optimization directions and hide capability-specific gains offset by regressions elsewhere, leaving promising directions underexplored. We introduce capability-driven evolution, which extends search guidance from overall performance to individual capability dimensions, preserving promising revisions and expanding exploration beyond the boundaries of holistic evolution. We propose PrisMem, which uses dependency-aware capability selection to prioritize targets with potential cross-capability benefits and history-guided diagnosis to refine capability specialists. Trace-guided integration compares evaluated programs on paired differential cases, using their behavioral differences to consolidate complementary gains into a unified memory program. Experiments show that PrisMem outperforms the strongest baselines by 10.54 and 7.83 percentage points on BEAM-1M and LongMemEval-M, respectively, demonstrating its effectiveness on million-token histories.

Figures & tables

Figure 1: Motivation for capability-driven evolution. Analysis of M ⋆ and EvolveMem evolution traces on BEAM. (a) Overall and maximum capability-level score changes for each revision. Many revisions improve at least one capability despite lower overall scores (upper-left region). (b) Among revisions that improve one capability (row), the percentage for which another capability’s score decreases (column). (c) Holistic evolution confines exploration to a limited region. (d) Capability-driven evolution enables broader exploration through explicit capability directions.

Capability	Memory requirement
Factual retrieval ( F )	Preserve and retrieve explicit facts, values, and other localized details.
Temporal tracking ( T )	Order events, determine recency, track states and knowledge updates.
Preference extraction ( P )	Recover and apply user preferences, instructions and constraints.
Multi-session synthesis ( M )	Aggregate and synthesize evidence across multiple sessions or events.
Adversarial ( A )	Abstain when memory lacks supporting evidence, or cautiously identify and resolve contradictory or misleading evidence.

Table 1: Five memory capabilities used in our experiments.

Figure 2: Overview of PrisMem . Stage I: Cold start quickly establishes a balanced program and initial evolution history. Stage II: Capability refinement uses dependency-aware capability selection and history-guided case selection to target capabilities and diagnose informative failures, respectively, refining their specialists through reflection and code revision. Stage III: Trace-guided integration compares the base program ( $\star$ ) and other capability specialists on paired differential cases, using their behavioral differences to guide integration into a unified memory program.

BEAM-1M
Method	Score (%) $\uparrow$						Avg. #Tokens (K) $\downarrow$
	Factual	Temporal	Preference	Multi-session	Adversarial	Overall
	retrieval	tracking	extraction	synthesis	Adversarial	Overall
Mem0	54.74 $\pm$ 3.06	46.51 $\pm$ 0.79	60.48 $\pm$ 2.39	38.02 $\pm$ 0.63	49.11 $\pm$ 1.24	48.95 $\pm$ 0.61	190.28 $\pm$ 0.20
A-MEM	56.21 $\pm$ 1.83	46.22 $\pm$ 0.62	65.39 $\pm$ 0.52	53.87 $\pm$ 0.86	41.13 $\pm$ 0.51	51.56 $\pm$ 0.62	428.95 $\pm$ 3.58
HippoRAG2	65.24 $\pm$ 1.26	47.06 $\pm$ 1.85	66.43 $\pm$ 1.57	44.83 $\pm$ 0.60	44.82 $\pm$ 2.28	51.86 $\pm$ 0.87	207.70 $\pm$ 0.19

Table 2: Performance on BEAM-1M and LongMemEval-M with Qwen3.8-27B. “Score” denotes the benchmark-provided LLM judge score, and “Avg. #Tokens” denotes the task LLM’s average token usage per question. All results are reported as mean

\pm

std. Best results are in bold .

Method	BEAM	LongMemEval
M ⋆	76.5 $\pm$ 7.1	202.3 $\pm$ 10.3
EvolveMem	69.2 $\pm$ 7.3	102.7 $\pm$ 8.4
PrisMem	73.7 $\pm$ 6.4	143.3 $\pm$ 9.6

Table 3: Evolution tokens (millions,

\downarrow

) with Qwen3.8-27B on two benchmarks.

Variant	Overall score	$\Delta$
Variant	(%) $\uparrow$	(pp)
Full PrisMem	63.39 $\pm$ 1.02	–
w/o capability evolution	53.24 $\pm$ 2.38	$-$ 10.15
w/o trace-guided integration	60.23 $\pm$ 1.86	$-$ 3.16
Round-robin target selection	60.84 $\pm$ 1.13	$-$ 2.55
Random case selection	58.13 $\pm$ 1.25	$-$ 5.26

Table 4: Ablations with Qwen3.8-27B on BEAM-1M.

\Delta

is relative to full PrisMem .

Panel of Figure 3 — Figure 3: Evolution trace analysis. (a) Gains of each Stage II iteration relative to program 05 obtained after cold start. Cell numbers are the creation iteration of the retained specialist for each capability; dots mark specialist updates. Scores improve intermittently across capabilities as refinement progresses. (b) Capability-wise scores of M ⋆ , EvolveMem, and PrisMem . Solid lines show the final selected programs, while dashed lines show the best capability-wise score reached across all iterations. PrisMem outperforms holistic evolution across capabilities, expanding the performance boundary, while its final integrated program closely matches its historical capability-wise best.

Appendix figures & tables9 assets

Supplementary material from the paper’s appendix.

Appendix

LoCoMo ( Maharana et al., 2024 )
Question type	Primary memory capabilities
Single-hop; Open-domain knowledge 1	Factual retrieval
Multi-hop	Multi-session synthesis
Temporal reasoning	Temporal tracking
Adversarial	Adversarial
LongMemEval ( Wu et al., 2025 )

Appendix

Table 5: Correspondence between benchmark question types and the memory capabilities used in our study. The entries describe primary requirements; individual questions may involve additional capabilities.

Changed component	Re-execution begins at	Reusable work
Memory-unit schema or extraction stage	Extraction	Segmentation
Constructor, index, or helpers	Indexing	Segmentation and extracted units
Query schema, planning, or retrieval	Query planning	Segmentation, extraction, and indexed memory
Answer instruction only	Answer generation	Built memory, query plan, and retrieved context

Appendix

Table 6: Re-execution boundaries for a candidate relative to its parent. For changes spanning several components, the earliest affected stage determines what can be reused.

Dataset	Split	F	T	P	M	A	Total	Avg. tokens (K)
BEAM-100K	Train	20	43	32	33	32	160	125.11
BEAM-100K	Validation	18	32	24	22	24	120	146.83
BEAM-1M	Test	70	210	140	140	140	700	1,089.57
LongMemEval-S	Train	19	17	11	9	4	60	109.40
LongMemEval-S	Validation	20	17	8	8	7	60	109.40
LongMemEval-M	Test	64	64	30	32	10	200	1,079.04

Appendix

Table 7: Sizes, primary capability composition, and average history length of each dataset split. F, T, P, M, and A denote factual retrieval, temporal tracking, preference extraction, multi-session synthesis, and adversarial, respectively. Average history length is measured with the Qwen3.8-27B tokenizer.

Stage II: Capability refinement
Parameter	Value
Case score threshold $\delta$	0.8
Refinement case budget $K$	5
Stage III: Integration
Capability-wise score degradation ratio $\theta$	0.1
Integration case budget	1 per capability

Appendix

Table 8: Parameter settings of PrisMem we used in experiments.

Method	Overall Score	Avg. #Tokens
	(%) $\uparrow$	(K) $\downarrow$
Mem0	57.77 $\pm$ 0.17	180.21 $\pm$ 0.12
A-MEM	60.92 $\pm$ 0.41	352.11 $\pm$ 0.25
HippoRAG2	57.73 $\pm$ 0.50	193.52 $\pm$ 0.18
SimpleMem	52.36 $\pm$ 0.55	121.39 $\pm$ 0.69
LightMem	46.68 $\pm$ 0.05	94.05 $\pm$ 0.19

Appendix

Table 9: Performance on BEAM with GPT-5.5.

Method	Evolved with Qwen3.8-27B (transfer)	Evolved with GPT-5.5 (target model)
PrisMem	67.25 $\pm$ 1.35	67.09 $\pm$ 1.79

Appendix

Table 10: Model transfer evaluated with GPT-5.5 on BEAM-1M. All programs are evolved on BEAM-100K. Overall scores (%) are reported as “mean

\pm

std” over three runs.

Method	Evolved on BEAM (transfer)	Evolved on LongMemEval (target dataset)
PrisMem	76.50 $\pm$ 1.67	75.50 $\pm$ 1.80

Appendix

Table 11: Dataset transfer evaluated with Qwen3.8-27B on LongMemEval-M. All programs are evolved with Qwen3.8-27B. Overall scores (%) are reported as “mean

\pm

std” over three runs.

Strategy	Overall Score (%)
Question-only router	57.74 $\pm$ 1.21
Oracle-label router	62.27 $\pm$ 1.14
PrisMem	63.39 $\pm$ 1.02

Appendix

Table 12: Routing versus integration on BEAM-1M with Qwen3.8-27B.

Evolution Trace	Failure diagnosis	Revision summary	Current best for
Stage I: Cold-start revisions
n05	A single flat query could not capture multiple semantic aspects of synthesis questions, causing retrieval to over-concentrate on one topic while missing complementary evidence scattered across sessions.	Decompose multi-aspect questions into focused sub-queries and retrieve each aspect independently with proportional budget allocation and cross-query deduplication. schema planning retrieval answering	M T A P F
Stage II: Capability-directed refinement
n05 $\to$ n06 Synthesis	Over-atomized memory units fragmented each logical item into multiple small facts, causing related evidence to appear as an unstructured list and hindering distinct-item counting and cross-session progression synthesis.	Add a topic label to each memory unit and group retrieved evidence by topic, so related facts are presented as coherent blocks for counting and multi-session synthesis. schema extraction indexing retrieval answering	F
n05 $\to$ n07 Temporal	Session-level temporal provenance was lost during indexing, so retrieved memories lacked discussion-time context and the model could not reliably distinguish statement order from event dates.	Add session date to each memory unit, recover it from the source session during indexing, and expose explicit date markers in retrieved evidence while preserving relevance-based ranking. schema extraction indexing retrieval answering helpers	F T P
n05 $\to$ n08 Adversarial	Memory units lacked speaker provenance, causing user-confirmed facts to be conflated with assistant recommendations or hypotheticals and obscuring genuine contradictions between user statements.	Add explicit speaker labels to memory units and propagate them through retrieval, then constrain answering to treat user-sourced units as direct evidence and surface conflicting user statements without resolution. schema extraction indexing retrieval answering	A

Appendix

Table 13: BEAM evolution traces with failure diagnoses, revision summaries, and current specialist. F : factual retrieval, T : temporal tracking, P : preference extraction, M : multi-session synthesis, A : adversarial.

Explore similar work

MemPro: Agentic Memory Systems as Evolvable Programs
May 30, 2026Qingshan Liu, Guoqing Wang, Wen Wu +5Agentic Memory Self-Evolving Agents
East China Normal University · Xiaohongshu Inc.
EvolveMem:Self-Evolving Memory Architecture via AutoResearch for LLM Agents
May 13, 2026Jiaqi Liu, Xinyu Ye, Peng Xia +4Alphaevolve Large Language Model Agents
UNC-Chapel Hill · UC Berkeley · UCSC
Mechanistic Attention Guidance for Agent Memory Refinement
Jul 20, 2026Yechao Hong, Haiquan Qiu, Yaqing Wang +1
Department of Electronic Engineering, Tsinghua University · Beijing Institute of Mathematical Sciences and Applications

May 30, 2026, cs.CLJ/K move · Enter open · S save

MemPro: Agentic Memory Systems as Evolvable Programs

Qingshan Liu, Guoqing Wang, Wen Wu, Jingqi Huang+4

East China Normal University · Xiaohongshu Inc.

Agentic Memory Self-Evolving Agents Production Agentic Systems

Open paper AskPDF arXiv

Long-horizon autonomous agents require memory systems to retain historical information, track evolving states, and reuse relevant knowledge beyond finite context windows. Existing agentic memory systems typically follow a memory construction-retrieval (MCR) pipeline, but often adapt mainly the memory bank while keeping the surrounding pipeline fixed after deployment. This fixed-pipeline design struggles to handle heterogeneous task-specific failure modes and can become misaligned with memory banks that evolve in scale and structure over time. To address these limitations, we propose MemPro, a system-level evolution framework that treats the entire MCR pipeline as an evolvable program rather than adapting only the memory bank or prompt text. MemPro maintains a version tree of runnable memory-system implementations, where an Evolving Agent iteratively selects promising versions, diagnoses recurring failures, and creates improved child versions through failure-mode-guided edit-debug refinement. Experiments on LongMemEval, LoCoMo, HotpotQA, and NarrativeQA show that MemPro consistently outperforms strong static and prompt-level evolving baselines within a few iterations, continues to improve with evolution, and achieves a favorable performance-cost trade-off. Code is available at https://github.com/wanghai673/MemPro.

Abstract

Explore similar work

MemPro: Agentic Memory Systems as Evolvable Programs

EvolveMem:Self-Evolving Memory Architecture via AutoResearch for LLM Agents

Mechanistic Attention Guidance for Agent Memory Refinement