Pith.

Computation and Language · arXiv:2611.17068 v1 · submitted 12 Nov 2026 · 92 references

Discovering Effective Agent Memory Designs through Program Evolution

Xinheng Ding, Seyyed Saeid Cheshmi, Ziyang Zhang, Tairan Chen, Wenya Xie, Yu Hou, Rui Zhang, Zirui Liu

The paper says

“Program evolution provides a promising approach to discover agent memory designs and study the choices underlying their effectiveness.”

more of the abstract

We introduce MAPE, a framework that evolves whole memory programs from minimal seeds through a standardized lifecycle interface and an iterative diagnosis-proposal loop. Experiments on LoCoMo and LongMemEval yield an absolute accuracy gain of up to 12% over manually designed baselines at 35% of the serving cost, with further evaluations supporting transfer across serving backbones and benchmarks. Empirical analyses and ablations identify and validate two recurring design patterns: retaining raw conversations as retrievable evidence preserves details omitted from extracted memories, while composing retrieved evidence according to query intent helps the answering model use it effectively. Transplanting the composition module into five existing memory systems improves accuracy across all five. Beyond these patterns, the search objective shapes where computation goes: optimizing serving cost alongside accuracy shifts computation from LLM-based memory construction to selective processing at query time while retaining competitive accuracy. Together, these findings offer empirical lessons for agent memory design. Code is available at https://github.com/ash-ding/MAPE.

from the abstract

Pith says

Machine review · deepseek-v4-flash · 13 Nov 2026 · what was read was not recorded · not checked by a person

It can be wrong. Challenge this review or reply to it.

Holds up, with conditions. A solid empirical paper on evolving agent memory systems; the evaluation-set selection concern is real but the paper's own transfer experiments partly mitigate it. This line is wrong

4 major objections · 5 minor

How often objections from reviews like this one are right: not yet measured

Are you an author? What you can do.

Not checked: figures, code, what the 92 references say.

The paper says, at §2.5 and Appendix B.2

The review did not quote this passage.

Pith says: Objection 1 of 4 · major

The final reported program for each setting is selected as the highest-evaluation-set-accuracy candidate among the shortlist, so the 80% evaluation set is used both for selection and for reporting. The paper states this explicitly (Appendix B.2), but it does not discuss how much of the reported margin could reflect selection on this particular evaluation partition.

more of objection 1

A genuinely unused held-out test split, a multiple-split estimate, or a bias-corrected selection analysis would be needed to establish the magnitude of the headline gains; the cross-benchmark transfer results provide indirect evidence of generalization but do not directly quantify within-benchmark selection bias.

The paper says, at Appendix C.1, Table 4

The review did not quote this passage.

Pith says: Objection 2 of 4 · major

The LoCoMo search partition contains only two conversations (conv-26 and conv-30, 233 search questions) out of ten. Because both the evolved programs and the reported transfer results are tied to this narrow search set and to the other eight conversations as the evaluation set, the 'up to 12%' headline may be quite sensitive to which two conversations were used for search.

more of objection 2

The paper should either provide a stability check across alternative splits or explicitly soften the claim from an estimate of expected gain to a result conditional on this fixed split.

The paper says, at §4.1, Tables 1 and 2 and Appendix D.1

The review did not quote this passage.

Pith says: Objection 3 of 4 · major

The cross-benchmark transfer comparisons are not fully controlled: on LoCoMo-to-LongMemEval transfer, the evolved programs (65.50% and 64.25%) do not beat RAG (66.00%), although the text in §4.1 says only that they outperform the five memory systems; on LongMemEval-to-LoCoMo transfer, MAPE-ALME's 75.90% exceeds all baselines.

more of objection 3

The claim 'selected systems outperform the evaluated human-designed baselines' is stated broadly in §1, yet the transfer tables show that this does not hold for RAG in one direction, and the abstract's phrasing could mislead readers into thinking the evolved programs beat every baseline everywhere. The claims in the abstract and introduction should be reworded to match exactly the comparisons that were actually run and won.

The paper says, at Appendix B.1, §2.4

The review did not quote this passage.

A classifier read this sentence of the paper as acknowledging the concern (probability 0.68):

“Evolutionary trajectories and resulting programs can vary across runs, and program selection and design interpretation involve empirical judgment.”

Pith says: Objection 4 of 4 · major

LongMemEval evolution is run only once per setting, and each run has 20 steps; the paper notes in Limitations that trajectories can vary across runs but does not report variance or repeated-run stability for the LongMemEval numbers. Given that the selection procedure picks the best evaluation-set candidate from a shortlist, a single run provides no indication of how much the 77.00% and 72.25% results would move under a different random seed or proposer trajectory.

5 minor objections

The paper says, at §3

Pith says: minor objection 1 of 5

The sentence 'LoCoMo is split by conversation, while LongMemEval-S is split by question instance together with its associated history' should clarify that LoCoMo's split inherits questions from the conversation partition, since Appendix C.1 states this explicitly; a reader of §3 alone might think questions were split directly.

The paper says, at §2.3, Eq. (2)

Pith says: minor objection 2 of 5

The cost formula is missing an explicit sum over lifecycle stages s that is described in prose; Eq. (2) defines per-stage cost, and the aggregation over stages appears only in the following sentence. Adding the stage index to the displayed sum would remove the ambiguity.

show 3 more

The paper says, at Appendix D.4, Figure 8

Pith says: minor objection 3 of 5

The estimated evidence completeness c-hat = a + (1-a)r is definitionally tied to accuracy, and the paper acknowledges this; the Spearman correlation between r and accuracy is the more informative quantity, but it is reported only in prose with no confidence interval or sample-size statement. Reporting the correlation with its uncertainty would strengthen the error-attribution claim.

The paper says, at Table 4

Pith says: minor objection 4 of 5

The paper states that adversarial questions are excluded from evaluation, but it does not explain what adversarial means on LoCoMo or whether the exclusion is standard for the benchmark; one sentence in Appendix C.1 defining the exclusion criterion would remove ambiguity.

The paper says, at §4.1

Pith says: minor objection 5 of 5

The ranking reversal between MAPE-ALCM and MAPE-PLCM under Qwen is attributed to MAPE-ALCM's failure to resolve relative dates, but Table 1 shows that MAPE-ALCM also drops on temporal questions while MAPE-PLCM improves; the explanation would be more convincing if the claimed 134-of-145 error pattern from Appendix D.1 were reported with the corresponding total temporal-error count for MAPE-PLCM.

What would settle it

Re-run the evaluation on a fresh held-out test set (e.g., an additional 10% of data not used in search or evaluation selection) for LoCoMo and LongMemEval. If the accuracy of the selected programs drops by more than 2-3% absolute compared to the reported evaluation-set accuracy, then the selection procedure has overfit to the evaluation partition and the headline gains are inflated.

from the skeptic's reading

How the review went

4 readings by deepseek-v4-flash · 13 Nov 2026

  1. Reader holds up, with conditions confidence moderate
  2. Skeptic partly agrees 1 min 41 s · 26,278 tokens read, 707 written
  3. Editor wrote the verdict above 59 s · 27,176 tokens read, 522 written
  4. Referee major revision: 4 major objections, 5 minor 20 s · 26,648 tokens read, 3,368 written

Not run on this paper: the simulated authors' defence.

Pith did not check this review's objections against the paper.

The review in full

The review in brief, machine-written

The paper introduces MAPE, a framework that evolves entire memory programs from minimal seeds using an LLM-driven search loop. The main empirical result is that evolved programs outperform several human-designed baselines on two QA benchmarks, often at lower serving cost. The paper also identifies two recurring design patterns—retaining raw conversations and query-routed composition—and shows that transplanting the composition module into existing systems improves accuracy across the board.

What is actually new: the combination of evolving all three lifecycle stages (build, retrieve, compose) from near-empty seeds, the joint accuracy–cost objective, and the systematic ablation/transplant analysis that isolates which design choices matter. The paper is well-written, the appendices are thorough, and the code is released. The cross-backbone and cross-benchmark transfer experiments are a genuine strength—they show the patterns generalize beyond the search setting.

The soft spot flagged by the stress test is real: the evaluation set (80% of data) is used for final program selection among shortlisted candidates. The paper acknowledges this involves empirical judgment but does not discuss overfitting risk. However, the concern is partly contained by (a) the transfer experiments, which test on a different benchmark without further evolution, and (b) the fact that the search set is held disjoint during evolution. A held-out test set would be cleaner, but the paper's design is not unusual for evolutionary search work. I would call this a moderate concern, not a fatal one.

Other soft spots: only one evolutionary run per setting on LongMemEval, no direct comparison with the closest prior work (MemPro, MemScope), and cost estimates use API rates for a locally served model. These are minor.

Who this is for: researchers working on agent memory, LLM-based program synthesis, or automated algorithm design. The paper deserves a serious referee—the empirical work is careful enough to engage with, and the design patterns are worth discussing even if the selection concern means the exact numbers should be taken with a grain of salt.

Recommendation: accept for peer review. The paper advances the subfield and the methodology is reproducible enough to verify.

The strongest objection

The paper's central claim that evolved programs outperform human-designed baselines rests on accuracy numbers reported on the evaluation set (80% of data). However, the final program for each setting is selected as the one with highest evaluation-set accuracy among shortlisted candidates (Appendix B.2). This means the evaluation set doubles as a selection set for the final reported program. While the search set (20%) is held separate during evolution, the evaluation set is used for both selection and reporting. If the evaluation set is not representative or if selection overfits to its particular composition, the reported gains may overstate performance on a truly held-out sample. The paper acknowledges that 'program selection and design interpretation involve empirical judgment' (Limitations), but does not discuss the risk of evaluation-set overfitting. The reader's weakest_assumption correctly identifies the fixed 20/80 split as a concern, but the deeper issue is that the 80% evaluation set is not a pure test set—it is used for final selection. A proper held-out test set (e.g., a further 10% split) would be needed to verify that the selected program's accuracy is not an artifact of selection on that particular partition. The paper's cross-benchmark transfer experiments provide some evidence of generalization, but they do not address within-benchmark overfitting to the evaluation set's specific questions.

The reader's weakest_assumption focuses on the representativeness of the 20/80 split and potential bias in evaluation-set selection. My concern is more specific: the evaluation set is used for final program selection (Appendix B.2: 'For each setting, we report the remaining program with the highest evaluation-set accuracy'). This creates a selection-on-the-test-set issue that is not fully addressed by the paper. The cross-benchmark transfer experiments (Tables 6, 7) provide some evidence of generalization, but they do not test within-benchmark generalization to a truly held-out sample. The paper's limitations mention that 'evolutionary trajectories and resulting programs can vary across runs' but do not discuss evaluation-set overfitting. I agree with the reader that the split is a concern, but I identify a more precise mechanism: the evaluation set serves as both selection and reporting set, which can inflate reported accuracy. My concrete test proposes a simple fix: hold out a further test set to verify that the selected program's accuracy is not an artifact of selection. I recommend keeping the verdict as CONDITIONAL because the concern is real but does not invalidate the paper's contributions; it merely suggests that the reported numbers should be interpreted with caution until a proper held-out test is performed.

The referee's report

If the results hold, the paper is a useful empirical contribution to agent memory design: it gives a concrete framework for whole-program evolution, full prompts and integration details, and it provides falsifiable, checkable claims about two design patterns (raw-conversation retention and query-routed composition) that are tested by ablation and by transplant into five third-party systems. The cost-aware analysis is a genuine addition to a literature that mostly optimizes accuracy alone, and the cross-benchmark/cross-backbone transfer results are a strong, non-obvious check on overfitting. The paper also ships complete prompt templates, documented subset partitions, and code, which raises the reproducibility standard for this type of work. The empirical pattern claims are plausible and useful even if the exact magnitude of the reported gains is partly contingent on the evaluation partitions and on the single evolutionary runs reported.

Pith's reading of the paper (a machine's summary, not the authors')

The paper claims that program evolution can replace much of manual memory engineering: starting from a seed program that stores nothing, an LLM-driven propose-and-diagnose loop evolves the entire memory pipeline as executable code, guided by measured accuracy and serving cost. On LoCoMo and LongMemEval, the selected programs outperform every evaluated human-designed baseline on their home benchmark, with gains up to 12 absolute accuracy points at as little as 35% of the serving cost, and the gains survive changing the serving model and moving to the other benchmark. The paper identifies two recurring design lessons: keep raw conversations as retrievable evidence because extracted summaries drop details, and route retrieved evidence through query-intent-specific scaffolds before answering. It also shows that the optimization objective shapes where computation goes, shifting from build-time LLM extraction to selective query-time processing when cost is optimized together with accuracy.

What carries the argument. The argument is carried by MAPE's standardized lifecycle interface, which exposes memory construction, retrieval, composition, and answering as four programmable stages, so the entire pipeline is open to revision under a shared evaluation protocol. Evolution loops through an LLM proposer that writes candidate programs with design hypotheses, a sanity check, evaluation on a fixed 20% search partition, and a diagnosis agent that attributes failures to build, retrieve, compose, or answer using execution traces. The joint objective uses the accuracy-cost Pareto frontier to guide proposals and retain cheap alternatives. The named artifact is Query-Routed Scaffolding (RQS), a compose-stage module that routes each query by intent and applies an intent-specific scaffold, such as date arithmetic for temporal questions or programmatic counting for aggregation, followed by a single deliberation call before the fixed answering model. RQS is validated by ablations and by transplanting it through a fixed adapter into five human-designed memory systems.

Core claim. On the paper's own terms, the central claim is that whole-program evolution from minimal seeds yields memory systems that outperform manually designed baselines and reveal transferable design patterns. All four selected systems, one per benchmark and objective, beat the evaluated human-designed baselines on their respective benchmarks, reaching 87.94% accuracy on LoCoMo and 77.00% on LongMemEval, with the joint objective delivering up to 12.01 absolute points over SimpleMem at about 35% of its serving cost. Two patterns recur across selected programs: retaining raw conversation turns as retrievable evidence and composing retrieved evidence through a named Query-Routed Scaffolding module before the fixed QA model answers. Ablations show that removing raw rounds or the scaffold costs several accuracy points on both benchmarks, and transplanting the scaffold into five existing memory systems improves accuracy in all five by 2.75 to 9.25 points. The authors also argue that the search objective acts as a design lever: accuracy-only search stacks operations and inflates cost for marginal gains, while joint accuracy-cost search keeps a branched lineage and ends with designs that compute selectively at query time.

The premise it rests on. The fixed 20/80 search and evaluation splits are representative enough that selecting the program with the highest evaluation-set accuracy does not inflate reported gains: LoCoMo search uses only two conversations (233 questions) and LongMemEval uses 100 of 500 instances, and an unrepresentative split could overstate margins and transfer results.

If the paper is right (Pith's inferences, not the authors')

Manual memory engineering can be partly replaced by objective-guided evolution: the strongest reported results come from programs that were searched, not designed, and they beat every evaluated hand-built baseline on both benchmarks.

The two recurring design lessons, raw-conversation retention and query-intent composition, survive across two benchmarks, two serving backbones, and, via RQS transplantation, five existing memory systems, so they are candidate general principles for conversational memory.

Optimizing cost alongside accuracy relocates computation from build-time LLM extraction to query-time selective processing, cutting serving calls by 40 to 90 percent while retaining competitive accuracy.

Cross-backbone robustness is not automatic: the accuracy-only program loses 36 points on temporal questions when the serving model changes, while the cost-aware program that resolves dates in code keeps its temporal accuracy.

Because every candidate carries a hypothesis, evaluation results, and failure attribution, the discovered designs yield inspectable lessons rather than opaque programs.

The raw-conversation result suggests practitioners can test the cheapest robustness fix first: keep an indexed raw-history side channel alongside extracted summaries, without changing the rest of the pipeline.

The RQS transplant gains being largest on temporal and counting questions implies that post-retrieval composition is the most neglected stage in current memory designs, and the ranking reversal across backbones suggests making temporal robustness an explicit search criterion would improve generalization.

The single-backbone reversal hints that evolution can overfit to one serving model's date-reasoning habits, so evolving under multiple backbones at once would be a natural extension of the same loop.

The cost-objective result implies that build-time reflection pipelines may not be the best use of LLM tokens for long-history conversational QA, and query-time selective computation is a cheap, testable alternative.

Check this review

Pith stored no record of this review's passes and took no fingerprint of it. Reviews made by code from 13 Nov 2026 on carry both.

How often reviews like this one are right

Reviews like this one: written by deepseek-v4-flash, where what was read was not recorded.

How often objections from reviews like this one are right: not yet measured

Every value above, in the review's public record (JSON)

Reviews by people

This review was written by a machine. No person has reviewed this paper on Pith yet. A scientist listed on Pith can review it for a fee, and their review is published here whatever it concludes.

Request a human review

How Pith checks papers · The full audit of this paper

The paper

Abstract

Program evolution provides a promising approach to discover agent memory designs and study the choices underlying their effectiveness. We introduce MAPE, a framework that evolves whole memory programs from minimal seeds through a standardized lifecycle interface and an iterative diagnosis-proposal loop. Experiments on LoCoMo and LongMemEval yield an absolute accuracy gain of up to 12% over manually designed baselines at 35% of the serving cost, with further evaluations supporting transfer across serving backbones and benchmarks. Empirical analyses and ablations identify and validate two recurring design patterns: retaining raw conversations as retrievable evidence preserves details omitted from extracted memories, while composing retrieved evidence according to query intent helps the answering model use it effectively. Transplanting the composition module into five existing memory systems improves accuracy across all five. Beyond these patterns, the search objective shapes where computation goes: optimizing serving cost alongside accuracy shifts computation from LLM-based memory construction to selective processing at query time while retaining competitive accuracy. Together, these findings offer empirical lessons for agent memory design. Code is available at https://github.com/ash-ding/MAPE.

The paper's own sentences, arranged, kept beside the review and separate from it. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. No sentence in this record was written by a model.

abstract compiled · 2026-11-14

The authors' main claim, as the classifier picked it out 0.62

“Together, these findings offer empirical lessons for agent memory design.”

How the main claim is stated: asserted 0.92

What the abstract itself states, as probabilities: claims first 0.05 · states limitation 0.14 · human participants 0.04 · controlled comparison 0.94 · randomized assignment 0.02 · states quantitative result 0.99 · compares with prior results 0.97 · mentions code or data availability 0.99

The abstract, sentence by sentence

This table labels the abstract only (arXiv listing abstract v1). The number is the probability the classifier gave the label.

SentenceRoleThe authors' own workStates the main conclusionStrength
Program evolution provides a promising approach to discover agent memory designs and study the choices underlying their effectiveness. background 0.99 0.10 0.07 hedged possibility 0.72
We introduce MAPE, a framework that evolves whole memory programs from minimal seeds through a standardized lifecycle interface and an iterative diagnosis-proposal loop. method 0.55 0.93 0.06 asserted 1.00
Experiments on LoCoMo and LongMemEval yield an absolute accuracy gain of up to 12% over manually designed baselines at 35% of the serving cost, with further evaluations supporting transfer across serving backbones and benchmarks. result 1.00 0.96 0.26 asserted 1.00
Empirical analyses and ablations identify and validate two recurring design patterns: retaining raw conversations as retrievable evidence preserves details omitted from extracted memories, while composing retrieved evidence according to query intent helps the answering model use it effectively. result 0.80 0.95 0.37 asserted 1.00
Transplanting the composition module into five existing memory systems improves accuracy across all five. result 0.99 0.95 0.09 asserted 1.00
Beyond these patterns, the search objective shapes where computation goes: optimizing serving cost alongside accuracy shifts computation from LLM-based memory construction to selective processing at query time while retaining competitive accuracy. result 0.59 0.93 0.15 asserted 0.99
Together, these findings offer empirical lessons for agent memory design. interpretation 0.97 0.49 0.32 asserted 0.92
Code is available at https://github.com/ash-ding/MAPE. contribution 0.83 0.44 0.02 asserted 0.85

What kind of paper the classifier reports this is: method or tool 0.99 · evidence from observation or experiment 0.99

What was examined, and what was not

Sources read

  • arXiv listing abstract v1

Not examined

  • full text
  • figures and tables
  • references and citing sentences
  • relations to other papers

Levels run

  • P abstract

Not run

  • S sections
  • T body sentences
  • C citations
  • R cross-paper relations

How the labels were made

Compiled by classifier-1.13.0 via TypeSafe API, question bank abstract-p1-20260923, 2026-11-14: 1 request, 6,194 input tokens.

Every quoted sentence on this record is the authors'. Every label is from Pith's closed vocabulary, with the probability shown beside it. This record as JSON

Figures (8)
Figure 1 of the paper
Figure 1. Overview of MAPE’s program evolution framework. Candidate programs are proposed using archived designs and diagnostic feedback, checked for execution errors, and evaluated on the search set. Their implementa￾tions, results, and diagnoses are retained to guide subsequent proposals. Evolution follows either an accuracy-only objective (MAPE-A) or joint accuracy–cost objectives (MAPE-P). Shortlisted programs undergo held-out evaluation for final selection. lowed by retrieval, composition, and answering for each associated query. Build. Bm maps the conversation history Hi to a persistent…
Figure 2 of the paper
Figure 2. Pipeline of MAPE-ALME, evolved under the accuracy-only objective on LongMemEval. Build retains both extracted atomic notes and raw conversations. Retrieve combines dense and sparse rankings through reciprocal rank fusion (RRF) and returns relevant notes and raw conversational evidence. Compose implements Query-Routed Scaffolding (RQS), which routes queries to type-specific scaffolds according to query intent, preparing context for the answering model. than their counterparts on LongMemEval and Lo￾CoMo, respectively. The larger reduction on Long￾MemEval is consistent with its…
Figure 3 of the paper
Figure 3. RQS transplanted from MAPE-ALME into five human-designed memory systems, evaluated on the LongMemEval evaluation set with Qwen3-30B-A3B￾Instruct-2507 as the serving backbone. Bold values indicate absolute accuracy gains in %. MemEval occurs for single-session-assistant ques￾tions, where accuracy drops by an absolute 20.00%. Category-level results are provided in
Figure 4 of the paper
Figure 4. A LoCoMo temporal example comparing MAPE-ALCM and MAPE-PLCM across serving backbones. The programs differ in the date representation and answer context passed to the QA model. Qwen answers are taken from held-out traces; GPT-4.1 mini answers are illustrative. Evidence and context excerpts are abbreviated. Method Temp Multi Know User Asst Pref Avg MAPE-ALME 80.19 58.49 88.89 89.29 93.33 54.17 77.00 w/o raw rounds 70.75 53.77 84.13 92.86 73.33 50.00 70.50 w/o RQS 60.38 56.60 80.95 94.64 95.56 50.00 70.75 w/o both 54.72 52.83 74.60 92.86 66.67 37.50 63.00
Figure 5 of the paper
Figure 5. LongMemEval search trajectories using Qwen3-30B-A3B-Instruct-2507 on the 100-question search set.
Figure 6 of the paper
Figure 6. LoCoMo search lineages using GPT-4.1 mini on the 233-question search set. Panels a and b show
Figure 7 of the paper
Figure 7. Evidence completeness among incorrect answers with the Qwen3-30B-A3B-Instruct-2507 serving
Figure 8 of the paper
Figure 8. Estimated overall evidence completeness and answer accuracy on LongMemEval in panel a and LoCoMo

Figures are the authors', reproduced from the paper.

References (92)
  1. [1] Zhengye Bao, Shengtao Zhang, Haokai Zhang, Shiqi Lou, Muning Wen, and Minnan Luo. 2026. https://arxiv.org/abs/2610.30755 Is complex agentic memory always needed? naive RAG is stronger than we thought . arXiv preprint arXiv:2610.30755 on PithPith's reviewarXiv
  2. [2] Tong Chen, Hongwei Wang, Sihao Chen, Wenhao Yu, Kaixin Ma, Xinran Zhao, Hongming Zhang, and Dong Yu. 2024. https://aclanthology.org/2024.emnlp-main.845/ Dense X retrieval: What retrieval granularity should we use? In Proceedings of the 2024 Conference on Empirical Methods in N... on Pith
  3. [3] Wenhu Chen, Xueguang Ma, Xinyi Wang, and William W. Cohen. 2023. https://openreview.net/pdf?id=YfZ4ZPt8zd Program of thoughts prompting: Disentangling computation from reasoning for numerical reasoning tasks . Transactions on Machine Learning Research on Pith
  4. [5] Prateek Chhikara, Dev Khant, Saket Aryan, Taranjeet Singh, and Deshraj Yadav. 2025. https://arxiv.org/abs/2504.19413 Mem0: Building production-ready AI agents with scalable long-term memory . arXiv preprint arXiv:2504.19413 on PithPith's reviewarXiv
  5. [6] Kalyanmoy Deb, Amrit Pratap, Sameer Agarwal, and T. Meyarivan. 2002. https://doi.org/10.1109/4235.996017 A fast and elitist multiobjective genetic algorithm: NSGA-II . IEEE Transactions on Evolutionary Computation, 6(2):182--197 on Pithdoi:10.1109/4235.996017
  6. [7] Xinle Deng, Ruobin Zhong, Hujin Peng, Xiaoben Lu, Yanzhe Wu, Guang Li, Buqiang Xu, Yunzhi Yao, Jizhan Fang, Haoliang Cao, Junjie Guo, Yuan Yuan, Ziqing Ma, Yuanqiang Yu, Rui Hu, Baohua Dong, Hangcheng Zhu, and Ningyu Zhang. 2026. https://arxiv.org/abs/2605.28732 MemTrace : Tra... on PithPith's reviewarXiv
  7. [8] Jizhan Fang, Xinle Deng, Haoming Xu, Ziyan Jiang, Yuqi Tang, Ziwen Xu, Shumin Deng, Yunzhi Yao, Mengru Wang, Shuofei Qiao, Huajun Chen, and Ningyu Zhang. 2025. https://arxiv.org/abs/2510.18866 LightMem : Lightweight and efficient memory-augmented generation . arXiv preprint ar... on PithPith's reviewarXiv
  8. [9] Luyu Gao, Aman Madaan, Shuyan Zhou, Uri Alon, Pengfei Liu, Yiming Yang, Jamie Callan, and Graham Neubig. 2023. https://proceedings.mlr.press/v202/gao23f.html PAL : Program-aided language models . In International Conference on Machine Learning, pages 10764--10799 on Pith
  9. [11] Patrick Lewis, Ethan Perez, Aleksandra Piktus, Fabio Petroni, Vladimir Karpukhin, Naman Goyal, Heinrich K \"u ttler, Mike Lewis, Wen-tau Yih, Tim Rockt \"a schel, Sebastian Riedel, and Douwe Kiela. 2020. https://proceedings.neurips.cc/paper/2020/hash/6b493230205f780e1bc26945df... on Pith
  10. [13] Fei Liu, Xialiang Tong, Mingxuan Yuan, Xi Lin, Fu Luo, Zhenkun Wang, Zhichao Lu, and Qingfu Zhang. 2024. https://proceedings.mlr.press/v235/liu24bs.html Evolution of heuristics: Towards efficient automatic algorithm design using large language model . In International Conferen... on Pith
  11. [14] Jiaqi Liu, Yaofeng Su, Peng Xia, Siwei Han, Zeyu Zheng, Cihang Xie, Mingyu Ding, and Huaxiu Yao. 2026 a . https://arxiv.org/abs/2601.02553 SimpleMem : Efficient lifelong memory for LLM agents . arXiv preprint arXiv:2601.02553 on PithPith's reviewarXiv
  12. [15] Jiaqi Liu, Xinyu Ye, Peng Xia, Zeyu Zheng, Cihang Xie, Mingyu Ding, and Huaxiu Yao. 2026 b . https://arxiv.org/abs/2605.13941 EvolveMem : Self-evolving memory architecture via AutoResearch for LLM agents . arXiv preprint arXiv:2605.13941 on PithPith's reviewarXiv
  13. [16] Qingshan Liu, Guoqing Wang, Wen Wu, Jingqi Huang, Xinqi Tao, Dejia Song, Jie Zhou, and Liang He. 2026 c . https://arxiv.org/abs/2606.00619 MemPro : Agentic memory systems as evolvable programs . arXiv preprint arXiv:2606.00619 on PithPith's reviewarXiv
  14. [17] Yecheng Jason Ma, William Liang, Guanzhi Wang, De-An Huang, Osbert Bastani, Dinesh Jayaraman, Yuke Zhu, Linxi Fan, and Anima Anandkumar. 2024. https://openreview.net/pdf?id=WbFBZ3bQUs Eureka: Human-level reward design via coding large language models . In International Confere... on Pith
  15. [18] Aman Madaan, Niket Tandon, Prakhar Gupta, Skyler Hallinan, Luyu Gao, Sarah Wiegreffe, Uri Alon, Nouha Dziri, Shrimai Prabhumoye, Yiming Yang, Shashank Gupta, Bodhisattwa Prasad Majumder, Katherine Hermann, Sean Welleck, Amir Yazdanbakhsh, and Peter Clark. 2023. https://proceed... on Pith
  16. [19] Adyasha Maharana, Dong-Ho Lee, Sergey Tulyakov, Mohit Bansal, Francesco Barbieri, and Yuwei Fang. 2024. https://aclanthology.org/2024.acl-long.747/ Evaluating very long-term conversational memory of LLM agents . In Proceedings of the 62nd Annual Meeting of the Association for ... on Pith
  17. [21] OpenAI . 2025. https://openai.com/index/gpt-4-1/ Introducing GPT-4.1 in the API on Pith
  18. [22] Charles Packer, Sarah Wooders, Kevin Lin, Vivian Fang, Shishir G. Patil, Ion Stoica, and Joseph E. Gonzalez. 2023. https://arxiv.org/abs/2310.08560 MemGPT : Towards LLMs as operating systems . arXiv preprint arXiv:2310.08560 on PithPith's reviewarXiv
  19. [23] Joon Sung Park, Joseph C. O'Brien, Carrie Jun Cai, Meredith Ringel Morris, Percy Liang, and Michael S. Bernstein. 2023. https://doi.org/10.1145/3586183.3606763 Generative agents: Interactive simulacra of human behavior . In Proceedings of the 36th Annual ACM Symposium on User ... on Pithdoi:10.1145/3586183.3606763
  20. [25] Stephen Robertson and Hugo Zaragoza. 2009. https://doi.org/10.1561/1500000019 The probabilistic relevance framework: BM25 and beyond . Foundations and Trends in Information Retrieval, 3(4):333--389 on Pithdoi:10.1561/1500000019
show 72 more
  1. [26] Bernardino Romera-Paredes, Mohammadamin Barekatain, Alexander Novikov, Matej Balog, M. Pawan Kumar, Emilien Dupont, Francisco J. R. Ruiz, Jordan S. Ellenberg, Pengming Wang, Omar Fawzi, Pushmeet Kohli, and Alhussein Fawzi. 2024. https://www.nature.com/articles/s41586-023-06924... on Pith
  2. [27] Noah Shinn, Federico Cassano, Ashwin Gopinath, Karthik Narasimhan, and Shunyu Yao. 2023. https://proceedings.neurips.cc/paper_files/paper/2023/hash/1b44b878bb782e6954cd888628510e90-Abstract-Conference.html Reflexion: Language agents with verbal reinforcement learning . In Adva... on Pith
  3. [28] Miao Su, Yucan Guo, Zixuan Li, Long Bai, Zhongni Hou, Xiaolong Jin, and Jiafeng Guo. 2026. https://arxiv.org/abs/2610.21170 MemScope : Evolving agent memory architectures through localized code repair . arXiv preprint arXiv:2610.21170 on PithPith's reviewarXiv
  4. [29] Lei Wang, Chen Ma, Xueyang Feng, Zeyu Zhang, Hao Yang, Jingsen Zhang, Zhiyuan Chen, Jiakai Tang, Xu Chen, Yankai Lin, Wayne Xin Zhao, Zhewei Wei, and Jirong Wen. 2024. https://doi.org/10.1007/s11704-024-40231-1 A survey on large language model based autonomous agents . Frontie... on Pithdoi:10.1007/s11704-024-40231-1
  5. [30] Di Wu, Hongwei Wang, Wenhao Yu, Yuwei Zhang, Kai-Wei Chang, and Dong Yu. 2025. https://openreview.net/pdf?id=pZiyCaVuti LongMemEval : Benchmarking chat assistants on long-term interactive memory . In International Conference on Learning Representations on Pith
  6. [33] Wujiang Xu, Zujie Liang, Kai Mei, Hang Gao, Juntao Tan, and Yongfeng Zhang. 2025. https://arxiv.org/abs/2502.12110 A-mem: Agentic memory for LLM agents . arXiv preprint arXiv:2502.12110 on PithPith's reviewarXiv
  7. [36] Jianing Yin, Tan Tang, and Yingcai Wu. 2026. https://arxiv.org/abs/2605.22411 DeferMem : Query-time evidence distillation via reinforcement learning for long-term agent memory . arXiv preprint arXiv:2605.22411 on PithPith's reviewarXiv
  8. [37] Jing Yu, Yibo Zhao, Jiaming Zhang, and Xiang Li. 2026. https://arxiv.org/abs/2607.22690 LazyMem : Retrieve broadly, construct selectively for efficient long-term agent memory . arXiv preprint arXiv:2607.22690 on PithPith's reviewarXiv
  9. [38] Guibin Zhang, Haotian Ren, Chong Zhan, Zhenhong Zhou, Junhao Wang, He Zhu, Wangchunshu Zhou, and Shuicheng Yan. 2026. https://arxiv.org/abs/2512.18746 MemEvolve : Meta-evolution of agent memory systems . In International Conference on Machine Learning. ArXiv:2512.18746 on PithPith's reviewarXiv
  10. [40] Lianmin Zheng, Wei-Lin Chiang, Ying Sheng, Siyuan Zhuang, Zhanghao Wu, Yonghao Zhuang, Zi Lin, Zhuohan Li, Dacheng Li, Eric P. Xing, Hao Zhang, Joseph E. Gonzalez, and Ion Stoica. 2023. https://proceedings.neurips.cc/paper_files/paper/2023/hash/91f18a1287b398d378ef22505bf41832... on Pith
  11. [41] Wanjun Zhong, Lianghong Guo, Qiqi Gao, He Ye, and Yanlin Wang. 2024. https://ojs.aaai.org/index.php/AAAI/article/view/29946 MemoryBank : Enhancing large language models with long-term memory . In Proceedings of the AAAI Conference on Artificial Intelligence, volume 38, pages 1... on Pith
  12. [42] Alfred V. Aho and Jeffrey D. Ullman , title =. 1972 on Pith
  13. [43] Publications Manual , year = "1983", publisher = on Pith
  14. [44] Ashok K. Chandra and Dexter C. Kozen and Larry J. Stockmeyer , year = "1981", title =. doi:10.1145/322234.322243 on Pithdoi:10.1145/322234.322243
  15. [45] Andrew, Galen and Gao, Jianfeng , booktitle=. Scalable training of on Pith
  16. [46] Dan Gusfield , title =. 1997 on Pith
  17. [47] Mohammad Sadegh Rasooli and Joel R. Tetreault , title =. Computing Research Repository , volume =. 2015 , url = on Pith
  18. [48] Ando, Rie Kubota and Zhang, Tong , Issn =. A Framework for Learning Predictive Structures from Multiple Tasks and Unlabeled Data , Volume =. Journal of Machine Learning Research , Month = dec, Numpages = on Pith
  19. [49] Packer, Charles and Wooders, Sarah and Lin, Kevin and Fang, Vivian and Patil, Shishir G. and Stoica, Ion and Gonzalez, Joseph E. , journal=. 2023 , url = on Pith
  20. [50] Chhikara, Prateek and Khant, Dev and Aryan, Saket and Singh, Taranjeet and Yadav, Deshraj , journal=. Mem0: Building Production-Ready. 2025 , url = on Pith
  21. [51] Xu, Wujiang and Liang, Zujie and Mei, Kai and Gao, Hang and Tan, Juntao and Zhang, Yongfeng , journal=. A-Mem: Agentic Memory for. 2025 , url = on Pith
  22. [52] Guti. Advances in Neural Information Processing Systems , year= on Pith
  23. [53] Mathematical Discoveries from Program Search with Large Language Models , author=. Nature , volume=. 2024 , url = on Pith
  24. [54] Novikov, Alexander and V. arXiv preprint arXiv:2506.13131 , year= on PithPith's reviewarXiv
  25. [55] Automated Design of Agentic Systems , author=. arXiv preprint arXiv:2408.08435 , year= on PithPith's reviewarXiv
  26. [56] Zhang, Jenny and Hu, Shengran and Lu, Cong and Lange, Robert and Clune, Jeff , journal=. Darwin. 2025 , url = on Pith
  27. [57] Agrawal, Lakshya A and Tan, Shangyin and Soylu, Dilara and Ziems, Noah and Khare, Rishi and Opsahl-Ong, Krista and Singhvi, Arnav and Shandilya, Herumb and Ryan, Michael J and Jiang, Meng and others , journal=. 2025 , url = on Pith
  28. [58] Zhang, Guibin and Ren, Haotian and Zhan, Chong and Zhou, Zhenhong and Wang, Junhao and Zhu, He and Zhou, Wangchunshu and Yan, Shuicheng , booktitle=. 2026 , note= on Pith
  29. [59] Liu, Qingshan and Wang, Guoqing and Wu, Wen and Huang, Jingqi and Tao, Xinqi and Song, Dejia and Zhou, Jie and He, Liang , journal=. 2026 , url = on Pith
  30. [60] Deb, Kalyanmoy and Pratap, Amrit and Agarwal, Sameer and Meyarivan, T. , journal=. A Fast and Elitist Multiobjective Genetic Algorithm:. 2002 , url = on Pith
  31. [61] Maharana, Adyasha and Lee, Dong-Ho and Tulyakov, Sergey and Bansal, Mohit and Barbieri, Francesco and Fang, Yuwei , booktitle=. Evaluating Very Long-Term Conversational Memory of. 2024 , address= on Pith
  32. [62] Wu, Di and Wang, Hongwei and Yu, Wenhao and Zhang, Yuwei and Chang, Kai-Wei and Yu, Dong , booktitle=. 2025 , url = on Pith
  33. [63] Generative Agents: Interactive Simulacra of Human Behavior , author=. Proceedings of the 36th Annual ACM Symposium on User Interface Software and Technology (UIST) , year= on Pith
  34. [64] Zhong, Wanjun and Guo, Lianghong and Gao, Qiqi and Ye, He and Wang, Yanlin , booktitle=. 2024 , url = on Pith
  35. [65] Zep: A Temporal Knowledge Graph Architecture for Agent Memory , author=. arXiv preprint arXiv:2501.13956 , year= on PithPith's reviewarXiv
  36. [66] Yan, Sikuan and Yang, Xiufeng and Huang, Zuchao and Nie, Ercong and Ding, Zifeng and Li, Zonggen and Ma, Xiaowen and Bi, Jinhe and Kersting, Kristian and Pan, Jeff Z. and Sch. arXiv preprint arXiv:2508.19828 , year= on PithPith's reviewarXiv
  37. [67] Mem- : Learning Memory Construction via Reinforcement Learning , author=. arXiv preprint arXiv:2509.25911 , year= on PithPith's reviewarXiv
  38. [68] Evolution of Heuristics: Towards Efficient Automatic Algorithm Design Using Large Language Model , author=. International Conference on Machine Learning , year= on Pith
  39. [69] Eureka: Human-Level Reward Design via Coding Large Language Models , author=. International Conference on Learning Representations , year= on Pith
  40. [70] Promptbreeder: Self-Referential Self-Improvement via Prompt Evolution , author=. arXiv preprint arXiv:2309.16797 , year= on PithPith's reviewarXiv
  41. [71] Zelikman, Eric and Lorch, Eliana and Mackey, Lester and Kalai, Adam Tauman , journal=. Self-Taught Optimizer (. 2023 , url = on Pith
  42. [72] Illuminating Search Spaces by Mapping Elites , author=. arXiv preprint arXiv:1504.04909 , year= on PithPith's reviewarXiv
  43. [73] Liu, Jiaqi and Ye, Xinyu and Xia, Peng and Zheng, Zeyu and Xie, Cihang and Ding, Mingyu and Yao, Huaxiu , journal=. 2026 , url = on Pith
  44. [74] A Survey on Large Language Model based Autonomous Agents , author=. Frontiers of Computer Science , volume=. 2024 , url = on Pith
  45. [75] The Rise and Potential of Large Language Model Based Agents: A Survey , author=. arXiv preprint arXiv:2309.07864 , year= on PithPith's reviewarXiv
  46. [76] A Survey on the Memory Mechanism of Large Language Model based Agents , author=. arXiv preprint arXiv:2404.13501 , year= on PithPith's reviewarXiv
  47. [77] Xin, Haidong and Li, Xinze and Liu, Zhenghao and Yan, Yukun and Wang, Shuo and Yang, Cheng and Gu, Yu and Yu, Ge and Sun, Maosong , booktitle=. 2026 , pages=. doi:10.18653/v1/2026.findings-acl.270 , url= on Pithdoi:10.18653/v1/2026.findings-acl.270
  48. [78] Deng, Xinle and Zhong, Ruobin and Peng, Hujin and Lu, Xiaoben and Wu, Yanzhe and Li, Guang and Xu, Buqiang and Yao, Yunzhi and Fang, Jizhan and Cao, Haoliang and Guo, Junjie and Yuan, Yuan and Ma, Ziqing and Yu, Yuanqiang and Hu, Rui and Dong, Baohua and Zhu, Hangcheng and Zha... on Pith
  49. [79] Xu, Xiucheng and Xu, Bingbing and Tian, Xueyun and Huang, Zihe and Chen, Rongxin and Li, Yunfan and Shen, Huawei , booktitle =. 2026 , month = jul, pages =. doi:10.18653/v1/2026.acl-long.534 , url = on Pithdoi:10.18653/v1/2026.acl-long.534
  50. [80] Fang, Jizhan and Deng, Xinle and Xu, Haoming and Jiang, Ziyan and Tang, Yuqi and Xu, Ziwen and Deng, Shumin and Yao, Yunzhi and Wang, Mengru and Qiao, Shuofei and Chen, Huajun and Zhang, Ningyu , journal=. 2025 , url= on Pith
  51. [81] Liu, Jiaqi and Su, Yaofeng and Xia, Peng and Han, Siwei and Zheng, Zeyu and Xie, Cihang and Ding, Mingyu and Yao, Huaxiu , journal=. 2026 , url= on Pith
  52. [82] Lewis, Patrick and Perez, Ethan and Piktus, Aleksandra and Petroni, Fabio and Karpukhin, Vladimir and Goyal, Naman and K. Retrieval-Augmented Generation for Knowledge-Intensive. Advances in Neural Information Processing Systems , volume=. 2020 , url = on Pith
  53. [83] Reflexion: Language Agents with Verbal Reinforcement Learning , author=. Advances in Neural Information Processing Systems , volume=. 2023 , url = on Pith
  54. [84] Self-Refine: Iterative Refinement with Self-Feedback , author=. Advances in Neural Information Processing Systems , volume=. 2023 , url = on Pith
  55. [85] Zheng, Lianmin and Chiang, Wei-Lin and Sheng, Ying and Zhuang, Siyuan and Wu, Zhanghao and Zhuang, Yonghao and Lin, Zi and Li, Zhuohan and Li, Dacheng and Xing, Eric P. and Zhang, Hao and Gonzalez, Joseph E. and Stoica, Ion , booktitle=. Judging. 2023 , url = on Pith
  56. [86] Robertson, Stephen and Zaragoza, Hugo , journal=. The Probabilistic Relevance Framework:. 2009 , url = on Pith
  57. [87] Cormack, Gordon V. and Clarke, Charles L. A. and Buettcher, Stefan , booktitle=. Reciprocal Rank Fusion Outperforms on Pith
  58. [88] Gao, Luyu and Madaan, Aman and Zhou, Shuyan and Alon, Uri and Liu, Pengfei and Yang, Yiming and Callan, Jamie and Neubig, Graham , booktitle=. 2023 , url = on Pith
  59. [89] Program of Thoughts Prompting: Disentangling Computation from Reasoning for Numerical Reasoning Tasks , author=. Transactions on Machine Learning Research , year= on Pith
  60. [90] Connecting Large Language Models with Evolutionary Algorithms Yields Powerful Prompt Optimizers , author=. International Conference on Learning Representations , year= on Pith
  61. [91] Large Language Models as Optimizers , author=. International Conference on Learning Representations , year= on Pith
  62. [92] General Agentic Memory Via Deep Research , author=. arXiv preprint arXiv:2511.18423 , year= on PitharXiv
  63. [93] Qwen3 Technical Report , author=. arXiv preprint arXiv:2505.09388 , year= on PithPith's reviewarXiv
  64. [94] Chen, Tong and Wang, Hongwei and Chen, Sihao and Yu, Wenhao and Ma, Kaixin and Zhao, Xinran and Zhang, Hongming and Yu, Dong , booktitle=. Dense. 2024 , url = on Pith
  65. [95] Introducing. 2025 , month = apr, url = on Pith
  66. [96] Yin, Jianing and Tang, Tan and Wu, Yingcai , journal =. 2026 , url = on Pith
  67. [97] Yu, Jing and Zhao, Yibo and Zhang, Jiaming and Li, Xiang , journal =. 2026 , url = on Pith
  68. [98] When Your Agent Opens the Chat App: Agent-Controlled Search over Raw Chat Logs Rivals Structured Memory , author =. arXiv preprint arXiv:2608.12888 , year = on PithPith's reviewarXiv
  69. [99] Bao, Zhengye and Zhang, Shengtao and Zhang, Haokai and Lou, Shiqi and Wen, Muning and Luo, Minnan , journal =. Is Complex Agentic Memory Always Needed? Naive. 2026 , url = on Pith
  70. [100] Memory Tax: Dissecting the Net Value of Agent Memory , author =. arXiv preprint arXiv:2610.34073 , year = on PithPith's reviewarXiv
  71. [101] Su, Miao and Guo, Yucan and Li, Zixuan and Bai, Long and Hou, Zhongni and Jin, Xiaolong and Guo, Jiafeng , journal=. 2026 , url= on Pith
  72. [102] Cheng, Zihao and Liu, Zeming and Shan, Yingyu and Wang, Xinyi and Zhu, Xiangrong and Ma, Yunpu and Wang, Hongru and Guo, Yuhang and Lin, Wei and Wang, Yunhong , booktitle=. 2026 , pages=. doi:10.18653/v1/2026.acl-long.952 , url= on Pithdoi:10.18653/v1/2026.acl-long.952

Challenges and discussion

Disagree with the review, or know something it missed? Leave a comment, or challenge this review formally: a challenge becomes a permanent entry in this paper's record, and its decision is published here.

No comments yet.

Sign in to comment