Pith. sign in

REVIEW 3 major objections 5 minor 9 cited by

This survey argues that efficiency in LLM-based agents is best understood as a cost–performance trade-off organized around three components—memory, tool learning, and planning—and that scattered methods converge on context compression, fewe

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · deepseek-v4-flash

2026-08-03 09:16 UTC pith:PBBSJDEN

load-bearing objection A useful organizing survey of agent-efficiency work with a clear taxonomy; the convergence claim needs a documented selection protocol to be fully trusted. the 3 major comments →

arxiv 2601.14192 v2 pith:PBBSJDEN submitted 2026-01-20 cs.AI cs.CL

Toward Efficient Agents: Memory, Tool learning, and Planning

Xiaofang Yang , Lijun Li , Heng Zhou , Tong Zhu , Xiaoye Qu , Yuchen Fan , Qianshan Wei , Rui Ye
show 7 more authors
Li Kang Yiran Qin Daizong Liu Qi Li Ning Ding Siheng Chen Jing Shao
This is my paper
classification cs.AI cs.CL
keywords efficient agentsagent memorytool learningplanningcost-performance trade-offtoken efficiencycontext compressionPareto frontier
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper attempts to put the scattered literature on LLM-agent efficiency under one roof. Its claim is that efficiency is not about shrinking the model but about optimizing the agent system—memory, tool learning, and planning—against a cost–performance trade-off: better results at a fixed budget, or lower cost at equal results, viewed through a Pareto frontier. Surveying many methods, the authors find convergent high-level principles: bound context through compression and management, use reinforcement-learning rewards to reduce tool invocations, and employ controlled search to trim planning effort. If this framing is right, evaluating agents should be inseparable from measuring their cost, and future efficiency work can be organized by which of the three components it targets. The paper also documents that current efficiency metrics are heterogeneous and often not directly comparable across papers.

Core claim

The paper's central discovery is organizational: the many efficiency techniques scattered across the agent literature are not a random collection. They fall into three functional components—memory (construction, management, access), tool learning (selection, calling, integrated reasoning), and planning (single-agent search and control plus multi-agent coordination)—and they converge on shared principles. The paper defines an efficient agent not as a smaller model but as an agentic system maximized for task success while minimizing resource consumption (tokens, latency, compute) across memory, tool usage, and planning. Efficiency itself is defined relationally: comparing effectiveness under a

What carries the argument

The load-bearing device is a three-part decomposition of the agent loop—memory, planning, and tool learning—paired with an additive cost model: total agent cost is approximately alpha times tokens plus tool-call costs plus memory-access costs plus retry costs. The cost model converts 'efficiency' from a vague property into a measurable budget, and the Pareto-frontier framing turns the goal into a two-objective problem. The taxonomy's lifecycle breakdowns—construct/manage/access memory, select/call/integrate tools, and budget/search/decompose planning—are the machinery that lets the survey map each published method to the specific cost it attacks.

Load-bearing premise

The paper's frame rests on the assumption that agent efficiency decomposes into three quasi-independent modules—memory, tool learning, and planning—with total cost roughly the sum of token, tool, memory, and retry costs; if those components interact significantly (for example, memory compression that causes extra tool calls), the taxonomy can misattribute where efficiency is gained or lost.

What would settle it

Run the same agent task battery under two memory-compression settings while holding tool and planning policies fixed. If tighter compression reduces memory tokens but measurably increases tool calls or retries—a cross-component cost—the additive model in Section 2.2 undercounts the true cost and the taxonomy's component boundaries misattribute efficiency.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Efficiency should be reported as a paired cost–performance measure—accuracy under a fixed token or latency budget, or cost at matched accuracy—rather than as standalone accuracy.
  • Memory work should be judged across the whole lifecycle, because construction, management, and access each create different bottlenecks: token growth, latency, and retrieval noise.
  • Tool learning is moving toward cost-aware training: rewards that penalize unnecessary tool calls, knapsack-style budget constraints, and parallel calling can shorten trajectories without hurting success.
  • Planning should be treated as resource-constrained control; adaptive deliberation, cost-aware tree search, and sparsified multi-agent communication all reduce the compute spent per solved task.
  • Because current efficiency metrics are defined inconsistently across papers, the field needs standardized reporting such as cost-of-pass—expected cost per successful completion—to make the trade-off measurable.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If the additive cost model is taken literally, the natural next experiment is to test cross-component interactions—for example, whether aggressive memory compression changes tool-call frequency or retry rates; a large interaction effect would push the field toward joint optimization rather than component-wise gains.
  • The convergence claim implies a predictive test: an agent whose memory, tool, and planning modules are each individually reported as efficient should also be efficient end-to-end; if that fails, the taxonomy needs revision.
  • The same cost–performance lens can be carried into multimodal agents, where re-encoding visual context at every step makes the memory–latency trade-off more severe than in text-only agents—an extension the paper lists as a challenge.
  • Cost-of-pass or similar economic metrics could become the unifying yardstick, converting the survey's Pareto framing into a single deployment-relevant number.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. This survey organizes recent work on efficiency in LLM-based agents into three components—memory, tool learning, and planning—and claims that diverse methods converge on shared principles such as context bounding, cost-aware tool invocation, and controlled search. It proposes a cost–performance trade-off framing (effectiveness at fixed cost, or cost at comparable effectiveness), reviews efficiency-oriented benchmarks and metrics, and discusses challenges and future directions. The paper is explicitly a survey and does not claim new algorithmic contributions.

Significance. If the convergence claim is accepted, the survey provides a useful organizing framework for a scattered literature. Its strengths include a broad taxonomy with detailed per-method categorization, structured tables (Tables 1–3), careful attention to cost-aware training signals in tool learning and planning, and a public GitHub repository for tracking work. The paper also honestly acknowledges metric heterogeneity and the lack of standardized evaluation, which raises the value of its proposed Pareto-frontier perspective. However, because the survey's central claim is an empirical generalization about field-wide convergence, the absence of a documented, reproducible selection protocol is a genuine limitation that affects the strength of the claim.

major comments (3)
  1. [Abstract, §1, §6, §8] The paper's central claim is that efficiency approaches 'frequently converge on shared high-level principles' (Abstract) and 'many seemingly different methods converge' (§8). This is an empirical generalization about the literature, but no search strategy, inclusion/exclusion criteria, coding protocol, or corpus list is provided. §1 states only 'we review a broad range of recent approaches' and §6 says 'we provide a high-level overview'; the GitHub repository is an awesome-list, not an auditable dataset. Without a reproducible sampling method, the convergence claim may reflect curation rather than field-wide convergence. Please add a methodology appendix describing the search and selection process, or soften the claim to describe patterns observed in the sampled works.
  2. [§3.4, §4.1, §5.2, Table 3] Several taxonomy entries come from the authors' own groups (e.g., AgentFold [175], ReSo [196], Diffthinker [35], memory survey [190]). This is not inherently problematic, but combined with the lack of a documented selection protocol, it increases the risk of selection bias in the convergence narrative. For the main claim to be load-bearing, the authors should either disclose a systematic search process or explicitly discuss how self-citations were included/excluded and whether their exclusion would change the reported patterns.
  3. [§7 and §6] Section 7 acknowledges that efficiency numbers are 'not directly comparable across papers' because metrics are defined differently (per query, per memory operation, per episode, etc.). This admission is important, but it creates tension with §6's goal of 'consolidating commonly reported efficiency metrics.' The paper would be strengthened by converting this tension into a concrete evaluation agenda—for example, proposing a minimal reporting template or standardizing cost-of-pass-style metrics—rather than leaving it as a future challenge.
minor comments (5)
  1. [§3.4] The method referred to as 'LatentMAS [204]' is listed in the references as 'Latent collaboration in multi-agent systems' by Zou et al. Please align the name used in the text with the reference title to avoid confusion.
  2. [§2.2] The cost model is given as an unnumbered display equation; it would be helpful to number it and to state explicitly that the indicator variables are coarse and that interactions (e.g., memory retrieval increasing tool calls) are not captured. The current text does note it is approximate, but a numbered equation would aid cross-referencing.
  3. [§6.1] The paragraph 'Efficiency Metrics in Memory Methods' repeats the same introductory sentence twice. Consider condensing to avoid redundancy.
  4. [§4.2] The names 'ToolChain*' and 'TableMind' appear in the text but not in the tables; ensure that all main methods discussed in the text are represented in the corresponding summary table for consistency.
  5. [Throughout] There are minor inconsistencies in capitalization and name formatting (e.g., 'ExpeL' vs 'Expel', 'MemOS' vs 'MemoryOS'). Please standardize these names, especially because MemoryOS [49] and MemOS [63] are different systems.

Circularity Check

0 steps flagged

No circular derivation: the survey's convergence claim is an inductive synthesis of external literature, and self-citations are illustrative, not load-bearing.

full rationale

This is a survey, not a derivation. The central claim—that agent-efficiency approaches 'frequently converge on shared high-level principles' (Abstract)—is an inductive generalization over cited external works, not a quantity derived from a fitted parameter. The cost model in Section 2.2 (Costagent ≈ αNtok + Itool·Costtool + Imem·Costmem + Iretry·Costretry) is an explicitly approximate decomposition that frames the survey's organizing lens; it is not used to predict any reported result. The taxonomy (memory, tool learning, planning) is a classification scheme applied to the literature, and the paper itself notes that some included methods are 'general memory modules for LLMs' reinterpreted 'from the view of efficient agent memory' (Section 3.1.1), which is a transparent editorial choice, not a concealed equivalence. Several entries are from the authors' own groups (AgentFold [175], ReSo [196], Diffthinker [35], and prior surveys [98, 182]), but each is used as one example among many to illustrate an existing category; no load-bearing argument depends on these self-citations. The absence of a documented search protocol (Section 1 only says 'we review a broad range of recent approaches') is a reproducibility limitation of an empirical generalization, not a circularity: it does not make any stated result true by construction. No equation, theorem, or fitted value is shown to reduce to its own input, so no circular step can be exhibited.

Axiom & Free-Parameter Ledger

0 free parameters · 4 axioms · 0 invented entities

No novel entities or fitted parameters are introduced. The survey's central contribution is an organizational frame, and that frame rests on a module decomposition, a simplified additive cost model, and an implicit representativeness assumption about the selected literature.

axioms (4)
  • domain assumption LLM-agent efficiency can be decomposed into three independent modules: memory, tool learning, and planning (Section 1, Figure 1).
    This is the survey's organizing taxonomy; the paper does not test whether the decomposition is complete or whether interactions between modules make it misleading.
  • domain assumption Efficiency is a cost-performance trade-off, evaluated either as effectiveness under fixed cost or cost at matched effectiveness, or via the Pareto frontier (Section 2.2, Section 6).
    Normative choice; the paper itself notes efficiency metrics are not standardized and numbers are not comparable across papers (Section 7).
  • ad hoc to paper Additive cost model: Cost_agent ≈ α N_tok + I_tool·Cost_tool + I_mem·Cost_mem + I_retry·Cost_retry (Section 2.2).
    Illustrative approximation; ignores overlaps (e.g., tool outputs stored in memory) and context-dependent latency.
  • domain assumption The reviewed works in Tables 1-3 are representative of the field.
    No systematic search or inclusion criteria are described; several entries are from the authors' own groups.

pith-pipeline@v1.3.0-alltime-deepseek · 40780 in / 12161 out tokens · 129321 ms · 2026-08-03T09:16:21.023548+00:00 · methodology

0 comments
read the original abstract

Recent years have witnessed increasing interest in extending large language models into agentic systems. While the effectiveness of agents has continued to improve, efficiency, which is crucial for real-world deployment, has often been overlooked. This paper therefore investigates efficiency from three core components of agents: memory, tool learning, and planning, considering costs such as latency, tokens, steps, etc. Aimed at conducting comprehensive research addressing the efficiency of the agentic system itself, we review a broad range of recent approaches that differ in implementation yet frequently converge on shared high-level principles including but not limited to bounding context via compression and management, designing reinforcement learning rewards to minimize tool invocation, and employing controlled search mechanisms to enhance efficiency, which we discuss in detail. Accordingly, we characterize efficiency in two complementary ways: comparing effectiveness under a fixed cost budget, and comparing cost at a comparable level of effectiveness. This trade-off can also be viewed through the Pareto frontier between effectiveness and cost. From this perspective, we also examine efficiency oriented benchmarks by summarizing evaluation protocols for these components and consolidating commonly reported efficiency metrics from both benchmark and methodological studies. Moreover, we discuss the key challenges and future directions, with the goal of providing promising insights.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 9 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Hackers or Hallucinators? A Comprehensive Analysis of LLM-Based Automated Penetration Testing

    cs.CR 2026-04 unverdicted novelty 8.0

    The first SoK on LLM-based AutoPT frameworks provides a six-dimension taxonomy of agent designs and a unified empirical benchmark evaluating 15 frameworks via over 10 billion tokens and 1,500 manually reviewed logs.

  2. LLM Agents Already Know When to Call Tools -- Even Without Reasoning

    cs.CL 2026-05 conditional novelty 7.0

    LLMs encode tool necessity in pre-generation hidden states at AUROC 0.89-0.96, enabling Probe&Prefill to reduce tool calls 48% with 1.7% accuracy loss, outperforming prompt and reasoning baselines.

  3. LLM Agents Already Know When to Call Tools -- Even Without Reasoning

    cs.CL 2026-05 accept novelty 7.0

    LLM agents encode tool necessity in pre-generation hidden states with high linear decodability (AUROC 0.89-0.96); Probe&Prefill uses this to reduce tool calls 48% with 1.7% accuracy loss.

  4. Tool-Adaptive LLM Reranker

    cs.IR 2026-07 conditional novelty 6.0

    Pointwise LLM reranking cast as an agentic MDP with language-preserving warm-up and asymmetric cost-aware RL yields adaptive tool use, SOTA NDCG, and high throughput.

  5. Are Rationales Necessary and Sufficient? Tuning LLMs for Explainable Misinformation Detection

    cs.CL 2026-05 unverdicted novelty 6.0

    LONSREX introduces a metric-based pipeline to identify necessary and sufficient rationales when creating training data for fine-tuning LLMs on explainable misinformation detection, addressing limitations of naive labe...

  6. HAGE: Harnessing Agentic Memory via RL-Driven Weighted Graph Evolution

    cs.AI 2026-05 unverdicted novelty 6.0

    HAGE proposes a trainable weighted graph memory framework with LLM intent classification, dynamic edge modulation, and RL optimization that improves long-horizon reasoning accuracy in agentic LLMs over static baselines.

  7. Select-then-Solve: Paradigm Routing as Inference-Time Optimization for LLM Agents

    cs.CL 2026-04 conditional novelty 6.0

    A learned embedding-based router selecting among six reasoning paradigms improves LLM agent accuracy from 47.6% to 53.1% on average, beating the best fixed paradigm by 2.8pp.

  8. READER: Robust Evidence-based Authorship Decoding via Extracted Representations

    cs.AI 2026-06 unverdicted novelty 4.0

    READER identifies source LLMs from variable-prompt generations at 31-42% single-response and 70-84% 50-response top-1 accuracy by proxy activation mapping and multi-query evidence accumulation, outperforming sentence ...

  9. OpenClaw and Ollama in Agentic AI: Toward Fully Autonomous and Scalable AI Agent Systems

    cs.AI 2026-04 reject novelty 3.0

    A review-plus-demo claiming agentic capabilities emerge from system integration, backed by a 15-task benchmark whose C1→C3 performance gap is largely built into the test design.

Reference graph

Works this paper leans on

209 extracted references · 3 canonical work pages · cited by 8 Pith papers

  1. [1]

    Lastras, Yara Rizk, and Pavan Kapanipathi

    Mayank Agarwal, Ibrahim Abdelaziz, Kinjal Basu, Merve Unuvar, Luis A. Lastras, Yara Rizk, and Pavan Kapanipathi. Toolrm: Outcome reward models for tool-calling large language models, 2025. URL https://arxiv.org/abs/2509.11963

    arXiv 2025

  2. [2]

    Rest meets react: Self-improvement for multi-step reasoning llm agent, 2023

    Renat Aksitov, Sobhan Miryoosefi, Zonglin Li, Daliang Li, Sheila Babayan, Kavya Kopparapu, Zachary Fisher, Ruiqi Guo, Sushant Prakash, Pranesh Srinivasan, Manzil Zaheer, Felix Yu, and Sanjiv Kumar. Rest meets react: Self-improvement for multi-step reasoning llm agent, 2023. URLhttps://arxiv. org/abs/2312.10003

    Pith/arXiv arXiv 2023

  3. [3]

    Smagdi: Socratic multi agent interaction graph distillation for efficient high accuracy reasoning.arXiv preprint arXiv:2511.05528, 2025

    Aayush Aluru, Myra Malik, Samarth Patankar, Spencer Kim, Kevin Zhu, Sean O’Brien, and Vasu Sharma. Smagdi: Socratic multi agent interaction graph distillation for efficient high accuracy reasoning.arXiv preprint arXiv:2511.05528, 2025

    arXiv 2025

  4. [4]

    Protip: Progressive tool retrieval improves planning, 2023

    Raviteja Anantha, Bortik Bandyopadhyay, Anirudh Kashi, Sayantan Mahinder, Andrew W Hill, and Srinivas Chappidi. Protip: Progressive tool retrieval improves planning, 2023. URLhttps://arxiv. org/abs/2312.10332

    Pith/arXiv arXiv 2023

  5. [5]

    Arigraph: Learning knowledge graph world models with episodic memory for llm agents.arXiv preprint arXiv:2407.04363, 2024

    Petr Anokhin, Nikita Semenov, Artyom Sorokin, Dmitry Evseev, Andrey Kravchenko, Mikhail Burtsev, and Evgeny Burnaev. Arigraph: Learning knowledge graph world models with episodic memory for llm agents.arXiv preprint arXiv:2407.04363, 2024. 35

    Pith/arXiv arXiv 2024

  6. [6]

    URLhttps://arxiv.org/abs/2506

    Victor Barres, Honghua Dong, Soham Ray, Xujie Si, and Karthik Narasimhan.τ2-bench: Evaluating conversational agents in a dual-control environment, 2025. URLhttps://arxiv.org/abs/2506. 07982

    2025

  7. [7]

    Titans: Learning to memorize at test time.arXiv preprint arXiv:2501.00663, 2024

    Ali Behrouz, Peilin Zhong, and Vahab Mirrokni. Titans: Learning to memorize at test time.arXiv preprint arXiv:2501.00663, 2024

    Pith/arXiv arXiv 2024

  8. [8]

    Autogpt+p: Affordance-based task planning with large language models, 2024

    Timo Birr, Christoph Pohl, Abdelrahman Younes, and Tamim Asfour. Autogpt+p: Affordance-based task planning with large language models, 2024. URLhttps://arxiv.org/abs/2402.10778

    Pith/arXiv arXiv 2024

  9. [9]

    Smurfs: Multi-agent system using context-efficient dfsdt for tool planning

    Junzhi Chen, Juhao Liang, and Benyou Wang. Smurfs: Multi-agent system using context-efficient dfsdt for tool planning. InProceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers), pages 3281–3298, 2025

    2025

  10. [10]

    Magdi: Structured distillation of multi-agent interaction graphs improves reasoning in smaller language models.arXiv preprint arXiv:2402.01620, 2024

    Justin Chih-Yao Chen, Swarnadeep Saha, Elias Stengel-Eskin, and Mohit Bansal. Magdi: Structured distillation of multi-agent interaction graphs improves reasoning in smaller language models.arXiv preprint arXiv:2402.01620, 2024

    Pith/arXiv arXiv 2024

  11. [11]

    Compress to impress: Unleashing the potential of compressive memory in real-world long-term conversations

    Nuo Chen, Hongguang Li, Jianhui Chang, Juhua Huang, Baoyuan Wang, and Jia Li. Compress to impress: Unleashing the potential of compressive memory in real-world long-term conversations. In Owen Rambow, Leo Wanner, Marianna Apidianaki, Hend Al-Khalifa, Barbara Di Eugenio, and Steven Schockaert, editors,Proceedings of the 31st International Conference on Com...

    2025

  12. [12]

    A 2fm: An adaptive agent foundation model for tool-aware hybrid reasoning, 2025

    Qianben Chen, Jingyi Cao, Jiayu Zhang, Tianrui Qin, Xiaowan Li, King Zhu, Dingfeng Shi, He Zhu, Minghao Liu, Xiaobo Liang, Xin Gui, Ge Zhang, Jian Yang, Yuchen Eleanor Jiang, and Wangchunshu Zhou. A 2fm: An adaptive agent foundation model for tool-aware hybrid reasoning, 2025. URL https://arxiv.org/abs/2510.12838

    arXiv 2025

  13. [13]

    T-eval: Evaluating the tool utilization capability of large language models step by step

    Zehui Chen, Weihua Du, Wenwei Zhang, Kuikun Liu, Jiangning Liu, Miao Zheng, Jingming Zhuo, Songyang Zhang, Dahua Lin, Kai Chen, et al. T-eval: Evaluating the tool utilization capability of large language models step by step. InProceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 9510–9529, 2024

    2024

  14. [14]

    Agent-flan: Designing data and methods of effective agent tuning for large language models, 2024

    Zehui Chen, Kuikun Liu, Qiuchen Wang, Wenwei Zhang, Jiangning Liu, Dahua Lin, Kai Chen, and Feng Zhao. Agent-flan: Designing data and methods of effective agent tuning for large language models, 2024. URLhttps://arxiv.org/abs/2403.12881

    Pith/arXiv arXiv 2024

  15. [15]

    Mem0: Building production-ready ai agents with scalable long-term memory.arXiv preprint arXiv:2504.19413, 2025

    Prateek Chhikara, Dev Khant, Saket Aryan, Taranjeet Singh, and Deshraj Yadav. Mem0: Building production-ready ai agents with scalable long-term memory.arXiv preprint arXiv:2504.19413, 2025

    Pith/arXiv arXiv 2025

  16. [16]

    Free-mad: Consensus-free multi-agent debate.arXiv preprint arXiv:2509.11035, 2025

    Yu Cui, Hang Fu, Haibin Zhang, Licheng Wang, and Cong Zuo. Free-mad: Consensus-free multi-agent debate.arXiv preprint arXiv:2509.11035, 2025

    Pith/arXiv arXiv 2025

  17. [17]

    Toolcoder: A systematic code-empowered tool learning framework for large language models, 2025

    Hanxing Ding, Shuchang Tao, Liang Pang, Zihao Wei, Jinyang Gao, Bolin Ding, Huawei Shen, and Xueqi Cheng. Toolcoder: A systematic code-empowered tool learning framework for large language models, 2025. URLhttps://arxiv.org/abs/2502.11404. 36

    Pith/arXiv arXiv 2025

  18. [18]

    AnyTool: Self-reflective, hierarchical agents for large-scale API calls

    Yu Du, Fangyun Wei, and Hongyang Zhang. AnyTool: Self-reflective, hierarchical agents for large-scale API calls. In Ruslan Salakhutdinov, Zico Kolter, Katherine Heller, Adrian Weller, Nuria Oliver, Jonathan Scarlett, and Felix Berkenkamp, editors,Proceedings of the 41st International Conference on Machine Learning, volume 235 ofProceedings of Machine Lear...

    2024

  19. [19]

    Tinyagent: Function calling at the edge

    Lutfi Eren Erdogan, Nicholas Lee, Siddharth Jha, Sehoon Kim, Ryan Tabrizi, Suhong Moon, Coleman Richard Charles Hooper, Gopala Anumanchipalli, Kurt Keutzer, and Amir Gholami. Tinyagent: Function calling at the edge. InProceedings of the 2024 Conference on Empirical Methods in Natural Language Processing: System Demonstrations, pages 80–88, 2024

    2024

  20. [20]

    Cost-of-pass: An economic framework for evaluating language models.arXiv preprint arXiv:2504.13359, 2025

    Mehmet Hamza Erol, Batu El, Mirac Suzgun, Mert Yuksekgonul, and James Zou. Cost-of-pass: An economic framework for evaluating language models.arXiv preprint arXiv:2504.13359, 2025

    arXiv 2025

  21. [21]

    Lightmem: Lightweight and efficient memory-augmented generation.arXiv preprint arXiv:2510.18866, 2025

    Jizhan Fang, Xinle Deng, Haoming Xu, Ziyan Jiang, Yuqi Tang, Ziwen Xu, Shumin Deng, Yunzhi Yao, Mengru Wang, Shuofei Qiao, et al. Lightmem: Lightweight and efficient memory-augmented generation.arXiv preprint arXiv:2510.18866, 2025

    Pith/arXiv arXiv 2025

  22. [22]

    Retool: Reinforcement learning for strategic tool use in llms, 2025

    Jiazhan Feng, Shijue Huang, Xingwei Qu, Ge Zhang, Yujia Qin, Baoquan Zhong, Chengquan Jiang, Jinxin Chi, and Wanjun Zhong. Retool: Reinforcement learning for strategic tool use in llms, 2025. URLhttps://arxiv.org/abs/2504.11536

    Pith/arXiv arXiv 2025

  23. [23]

    Budgetmlagent: A cost- effective llm multi-agent system for automating machine learning tasks, 2025

    Shubham Gandhi, Manasi Patwardhan, Lovekesh Vig, and Gautam Shroff. Budgetmlagent: A cost- effective llm multi-agent system for automating machine learning tasks, 2025. URLhttps://arxiv. org/abs/2411.07464

    Pith/arXiv arXiv 2025

  24. [24]

    Memory sharing for large language model based agents.arXiv preprint arXiv:2404.09982, 2024

    Hang Gao and Yongfeng Zhang. Memory sharing for large language model based agents.arXiv preprint arXiv:2404.09982, 2024

    arXiv 2024

  25. [25]

    Efficient tool use with chain-of- abstraction reasoning, 2025

    Silin Gao, Jane Dwivedi-Yu, Ping Yu, Xiaoqing Ellen Tan, Ramakanth Pasunuru, Olga Golovneva, Koustuv Sinha, Asli Celikyilmaz, Antoine Bosselut, and Tianlu Wang. Efficient tool use with chain-of- abstraction reasoning, 2025. URLhttps://arxiv.org/abs/2401.17464

    Pith/arXiv arXiv 2025

  26. [26]

    Mcp-radar: A multi-dimensional benchmark for evaluating tool use capabilities in large language models, 2025

    Xuanqi Gao, Siyi Xie, Juan Zhai, Shiqing Ma, and Chao Shen. Mcp-radar: A multi-dimensional benchmark for evaluating tool use capabilities in large language models, 2025. URLhttps:// arxiv.org/abs/2505.16700

    arXiv 2025

  27. [27]

    Learning to forget: Continual prediction with lstm.Neural computation, 12(10):2451–2471, 2000

    Felix A Gers, Jürgen Schmidhuber, and Fred Cummins. Learning to forget: Continual prediction with lstm.Neural computation, 12(10):2451–2471, 2000

    2000

  28. [28]

    Anna Goldie, Azalia Mirhoseini, Hao Zhou, Irene Cai, and Christopher D. Manning. Synthetic data generation & multi-step rl for reasoning & tool use, 2025. URLhttps://arxiv.org/abs/2504. 04736

    2025

  29. [29]

    Towards an ai co- scientist: A multi-agent system for scientific discovery.arXiv preprint arXiv:2502.18864, page 3, 2025

    J Gottweis, WH Weng, A Daryin, T Tu, A Palepu, P Sirkovic, and V Natarajan. Towards an ai co- scientist: A multi-agent system for scientific discovery.arXiv preprint arXiv:2502.18864, page 3, 2025

    Pith/arXiv arXiv 2025

  30. [30]

    Stabletoolbench: Towards stable large-scale benchmarking on tool learning of large language models, 2025

    Zhicheng Guo, Sijie Cheng, Hao Wang, Shihao Liang, Yujia Qin, Peng Li, Zhiyuan Liu, Maosong Sun, and Yang Liu. Stabletoolbench: Towards stable large-scale benchmarking on tool learning of large language models, 2025. URLhttps://arxiv.org/abs/2403.07714. 37

    Pith/arXiv arXiv 2025

  31. [31]

    Legomem: Modular procedural memory for multi-agent llm systems for workflow automation

    Dongge Han, Camille Couturier, Daniel Madrigal Diaz, Xuchao Zhang, Victor Rühle, and Saravan Raj- mohan. Legomem: Modular procedural memory for multi-agent llm systems for workflow automation. arXiv preprint arXiv:2510.04851, 2025

    arXiv 2025

  32. [32]

    Nestools: A dataset for evaluating nested tool learning abilities of large language models, 2025

    Han Han, Tong Zhu, Xiang Zhang, Mengsong Wu, Hao Xiong, and Wenliang Chen. Nestools: A dataset for evaluating nested tool learning abilities of large language models, 2025. URLhttps: //arxiv.org/abs/2410.11805

    Pith/arXiv arXiv 2025

  33. [33]

    Toolkengpt: Augmenting frozen language models with massive tools via tool embeddings

    Shibo Hao, Tianyang Liu, Zhen Wang, and Zhiting Hu. Toolkengpt: Augmenting frozen language models with massive tools via tool embeddings. In A. Oh, T. Naumann, A. Globerson, K. Saenko, M. Hardt, and S. Levine, editors,Advances in Neural Information Processing Systems, volume 36, pages 45870–45894. Curran Associates, Inc., 2023

    2023

  34. [34]

    WebVoyager: Building an end-to-end web agent with large multimodal models

    Hongliang He, Wenlin Yao, Kaixin Ma, Wenhao Yu, Yong Dai, Hongming Zhang, Zhenzhong Lan, and Dong Yu. WebVoyager: Building an end-to-end web agent with large multimodal models. In Lun-Wei Ku, Andre Martins, and Vivek Srikumar, editors,Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 686...

    doi:10.18653/v1/2024.acl-long 2024

  35. [35]

    Diffthinker: Towards generative multimodal reasoning with diffusion models.arXiv preprint arXiv:2512.24165, 2025

    Zefeng He, Xiaoye Qu, Yafu Li, Tong Zhu, Siyuan Huang, and Yu Cheng. Diffthinker: Towards generative multimodal reasoning with diffusion models.arXiv preprint arXiv:2512.24165, 2025

    arXiv 2025

  36. [36]

    Planning without search: Refining frontier llms with offline goal-conditioned rl, 2025

    Joey Hong, Anca Dragan, and Sergey Levine. Planning without search: Refining frontier llms with offline goal-conditioned rl, 2025. URLhttps://arxiv.org/abs/2505.18098

    arXiv 2025

  37. [37]

    Metagpt: Meta programming for a multi- agent collaborative framework

    Sirui Hong, Mingchen Zhuge, Jonathan Chen, Xiawu Zheng, Yuheng Cheng, Jinlin Wang, Ceyao Zhang, Zili Wang, Steven Ka Shing Yau, Zijuan Lin, et al. Metagpt: Meta programming for a multi- agent collaborative framework. InThe Twelfth International Conference on Learning Representations, 2023

    2023

  38. [38]

    Cogagent: Avisual language modelforgui agents

    Wenyi Hong, Weihan Wang, Qingsong Lv, Jiazheng Xu, Wenmeng Yu, Junhui Ji, Yan Wang, Zihan Wang, Yuxiao Dong, Ming Ding, etal. Cogagent: Avisual language modelforgui agents. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 14281–14290, 2024

    2024

  39. [39]

    my agent understands me better

    Yuki Hou, Haruki Tamoto, and Homei Miyashita. " my agent understands me better": Integrating dynamic human-like memory recall and consolidation in llm-based agents. InExtended Abstracts of the CHI Conference on Human Factors in Computing Systems, pages 1–7, 2024

    2024

  40. [40]

    Hiagent: Hierar- chical working memory management for solving long-horizon agent tasks with large language model

    Mengkang Hu, Tianxing Chen, Qiguang Chen, Yao Mu, Wenqi Shao, and Ping Luo. Hiagent: Hierar- chical working memory management for solving long-horizon agent tasks with large language model. InProceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 32779–32798, 2025

    2025

  41. [41]

    Planning, creation, usage: Benchmarking llms for comprehensive tool utilization in real-world complex scenarios, 2024

    Shijue Huang, Wanjun Zhong, Jianqiao Lu, Qi Zhu, Jiahui Gao, Weiwen Liu, Yutai Hou, Xingshan Zeng, Yasheng Wang, Lifeng Shang, Xin Jiang, Ruifeng Xu, and Qun Liu. Planning, creation, usage: Benchmarking llms for comprehensive tool utilization in real-world complex scenarios, 2024. URL https://arxiv.org/abs/2401.17167

    Pith/arXiv arXiv 2024

  42. [42]

    Metatool benchmark for large language models: Deciding whether to use tools and which to use, 2024

    Yue Huang, Jiawen Shi, Yuan Li, Chenrui Fan, Siyuan Wu, Qihui Zhang, Yixin Liu, Pan Zhou, Yao Wan, Neil Zhenqiang Gong, and Lichao Sun. Metatool benchmark for large language models: Deciding whether to use tools and which to use, 2024. URLhttps://arxiv.org/abs/2310.03128. 38

    Pith/arXiv arXiv 2024

  43. [43]

    Reinforced internal-external knowledge synergistic reasoning for efficient adaptive search agent.arXiv preprint arXiv:2505.07596, 2025

    Ziyang Huang, Xiaowei Yuan, Yiming Ju, Jun Zhao, and Kang Liu. Reinforced internal-external knowledge synergistic reasoning for efficient adaptive search agent.arXiv preprint arXiv:2505.07596, 2025

    Pith/arXiv arXiv 2025

  44. [44]

    Gpt-4o system card.arXiv preprint arXiv:2410.21276, 2024

    Aaron Hurst, Adam Lerer, Adam P Goucher, Adam Perelman, Aditya Ramesh, Aidan Clark, AJ Ostrow, Akila Welihinda, Alan Hayes, Alec Radford, et al. Gpt-4o system card.arXiv preprint arXiv:2410.21276, 2024

    Pith/arXiv arXiv 2024

  45. [45]

    Tablemind: An autonomousprogrammaticagentfortool-augmentedtablereasoning.arXivpreprintarXiv:2509.06278, 2025

    Chuang Jiang, Mingyue Cheng, Xiaoyu Tao, Qingyang Mao, Jie Ouyang, and Qi Liu. Tablemind: An autonomousprogrammaticagentfortool-augmentedtablereasoning.arXivpreprintarXiv:2509.06278, 2025

    Pith/arXiv arXiv 2025

  46. [46]

    KG-agent: An efficient autonomous agent framework for complex reasoning over knowledge graph

    Jinhao Jiang, Kun Zhou, Wayne Xin Zhao, Yang Song, Chen Zhu, Hengshu Zhu, and Ji-Rong Wen. KG-agent: An efficient autonomous agent framework for complex reasoning over knowledge graph. In Wanxiang Che, Joyce Nabende, Ekaterina Shutova, and Mohammad Taher Pilehvar, editors, Proceedings of the 63rd Annual Meeting of the Association for Computational Linguis...

    2025

  47. [47]

    SWE-bench: Can language models resolve real-world github issues? InThe Twelfth International Conference on Learning Representations, 2024

    Carlos E Jimenez, John Yang, Alexander Wettig, Shunyu Yao, Kexin Pei, Ofir Press, and Karthik R Narasimhan. SWE-bench: Can language models resolve real-world github issues? InThe Twelfth International Conference on Learning Representations, 2024. URLhttps://openreview.net/ forum?id=VTF8yNQM66

    2024

  48. [48]

    Benchmark for planning and control with large language model agents: Blocksworld with model context protocol.arXiv preprint arXiv:2512.03955, 2025

    Niklas Jobs, Luis Miguel Vieira da Silva, Jayanth Somashekaraiah, Maximilian Weigand, David Kube, and Felix Gehlhoff. Benchmark for planning and control with large language model agents: Blocksworld with model context protocol.arXiv preprint arXiv:2512.03955, 2025

    arXiv 2025

  49. [49]

    Memory OS of AI agent

    Jiazheng Kang, Mingming Ji, Zhe Zhao, and Ting Bai. Memory OS of AI agent. In Christos Christodoulopoulos, Tanmoy Chakraborty, Carolyn Rose, and Violet Peng, editors,Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, pages 25961–25970, Suzhou, China, November 2025. Association for Computational Linguistics. ISBN 979-8-...

    doi:10.18653/v1/2025.emnlp-main.1318 2025

  50. [50]

    Distilling llm agent into small models with retrieval and code tools, 2025

    Minki Kang, Jongwon Jeong, Seanie Lee, Jaewoong Cho, and Sung Ju Hwang. Distilling llm agent into small models with retrieval and code tools, 2025. URLhttps://arxiv.org/abs/2505.17612

    arXiv 2025

  51. [51]

    An llm compiler for parallel function calling

    Sehoon Kim, Suhong Moon, Ryan Tabrizi, Nicholas Lee, Michael W Mahoney, Kurt Keutzer, and Amir Gholami. An llm compiler for parallel function calling. InForty-first International Conference on Machine Learning, 2024

    2024

  52. [52]

    Dai, Jakob Uszkoreit, Quoc Le, and Slav Petrov

    Tom Kwiatkowski, Jennimaria Palomaki, Olivia Redfield, Michael Collins, Ankur Parikh, Chris Alberti, Danielle Epstein, Illia Polosukhin, Jacob Devlin, Kenton Lee, Kristina Toutanova, Llion Jones, Matthew Kelcey, Ming-Wei Chang, Andrew M. Dai, Jakob Uszkoreit, Quoc Le, and Slav Petrov. Natural questions: A benchmark for question answering research.Transact...

    doi:10.1162/tacl_a_00276 2019

  53. [53]

    Convolutional networks for images, speech, and time series.The handbook of brain theory and neural networks, 1998

    Yann LeCun and Yoshua Bengio. Convolutional networks for images, speech, and time series.The handbook of brain theory and neural networks, 1998

    1998

  54. [54]

    A human-inspired reading agent with gist memory of very long contexts

    Kuang-Huei Lee, Xinyun Chen, Hiroki Furuta, John Canny, and Ian Fischer. A human-inspired reading agent with gist memory of very long contexts. InProceedings of the 41st International Conference on Machine Learning, pages 26396–26415, 2024

    2024

  55. [55]

    D-smart: Enhancingllmdialogueconsistencyviadynamicstructured memory and reasoning tree.arXiv preprint arXiv:2510.13363, 2025

    XiangLei, QinLi, andMinZhang. D-smart: Enhancingllmdialogueconsistencyviadynamicstructured memory and reasoning tree.arXiv preprint arXiv:2510.13363, 2025

    arXiv 2025

  56. [56]

    Camel: Commu- nicative agents for" mind" exploration of large language model society.Advances in Neural Information Processing Systems, 36:51991–52008, 2023

    Guohao Li, Hasan Hammoud, Hani Itani, Dmitrii Khizbullin, and Bernard Ghanem. Camel: Commu- nicative agents for" mind" exploration of large language model society.Advances in Neural Information Processing Systems, 36:51991–52008, 2023

    2023

  57. [57]

    Hello again! llm-powered personalized agent for long-term dialogue

    Hao Li, Chenghao Yang, An Zhang, Yang Deng, Xiang Wang, and Tat-Seng Chua. Hello again! llm-powered personalized agent for long-term dialogue. InProceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers), pages 5259–5276, 2025

    2025

  58. [58]

    Seek in the dark: Reasoning via test-time instance-level policy gradient in latent space, 2025

    Hengli Li, Chenxi Li, Tong Wu, Xuekai Zhu, Yuxuan Wang, Zhaoxin Yu, Eric Hanchen Jiang, Song- Chun Zhu, Zixia Jia, Ying Nian Wu, and Zilong Zheng. Seek in the dark: Reasoning via test-time instance-level policy gradient in latent space, 2025. URLhttps://arxiv.org/abs/2505.13308

    arXiv 2025

  59. [59]

    API-bank: A comprehensive benchmark for tool-augmented LLMs

    Minghao Li, Yingxiu Zhao, Bowen Yu, Feifan Song, Hangyu Li, Haiyang Yu, Zhoujun Li, Fei Huang, and Yongbin Li. API-bank: A comprehensive benchmark for tool-augmented LLMs. In Houda Bouamor, Juan Pino, and Kalika Bali, editors,Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 3102–3116, Singapore, December 2023. ...

    doi:10.18653/v1/2023.emnlp-main.187 2023

  60. [60]

    GraphReader: Building graph-based agent to enhance long-context abilities of large language models

    Shilong Li, Yancheng He, Hangyu Guo, Xingyuan Bu, Ge Bai, Jie Liu, Jiaheng Liu, Xingwei Qu, Yangguang Li, Wanli Ouyang, Wenbo Su, and Bo Zheng. GraphReader: Building graph-based agent to enhance long-context abilities of large language models. In Yaser Al-Onaizan, Mohit Bansal, and Yun-Nung Chen, editors,Findings of the Association for Computational Lingu...

    doi:10.18653/v1/2024.f 2024

  61. [61]

    Optimus-1: Hybrid multimodal memory empowered agents excel in long-horizon tasks.Advances in neural information processing systems, 37:49881–49913, 2024

    Zaijing Li, Yuquan Xie, Rui Shao, Gongwei Chen, Dongmei Jiang, and Liqiang Nie. Optimus-1: Hybrid multimodal memory empowered agents excel in long-horizon tasks.Advances in neural information processing systems, 37:49881–49913, 2024

    2024

  62. [62]

    Encouraging good processes without the need for good answers: Reinforcement learning for llm agent planning, 2025

    Zhiwei Li, Yong Hu, and Wenqing Wang. Encouraging good processes without the need for good answers: Reinforcement learning for llm agent planning, 2025. URLhttps://arxiv.org/abs/ 2508.19598

    Pith/arXiv arXiv 2025

  63. [63]

    Memos: A memory os for ai system.arXiv preprint arXiv:2507.03724, 2025

    Zhiyu Li, Shichao Song, Chenyang Xi, Hanyu Wang, Chen Tang, Simin Niu, Ding Chen, Jiawei Yang, Chunyu Li, Qingchen Yu, et al. Memos: A memory os for ai system.arXiv preprint arXiv:2507.03724, 2025. 40

    Pith/arXiv arXiv 2025

  64. [64]

    Swiftsage: A generative agent with fast and slow thinking for complex interactive tasks, 2023

    Bill Yuchen Lin, Yicheng Fu, Karina Yang, Faeze Brahman, Shiyu Huang, Chandra Bhagavatula, Prithviraj Ammanabrolu, Yejin Choi, and Xiang Ren. Swiftsage: A generative agent with fast and slow thinking for complex interactive tasks, 2023. URLhttps://arxiv.org/abs/2305.17390

    Pith/arXiv arXiv 2023

  65. [65]

    Pierrehumbert

    Fangru Lin, Emanuele La Malfa, Valentin Hofmann, Elle Michelle Yang, Anthony Cohn, and Janet B. Pierrehumbert. Graph-enhanced large language models in asynchronous plan reasoning, 2024. URL https://arxiv.org/abs/2402.02805

    Pith/arXiv arXiv 2024

  66. [66]

    Stop wasting your tokens: Towards efficient runtime multi-agent systems.arXiv preprint arXiv:2510.26585, 2025

    Fulin Lin, Shaowen Chen, Ruishan Fang, Hongwei Wang, and Tao Lin. Stop wasting your tokens: Towards efficient runtime multi-agent systems.arXiv preprint arXiv:2510.26585, 2025

    arXiv 2025

  67. [67]

    Qlass: Boosting language agent inference via q-guided stepwise search.arXiv preprint arXiv:2502.02584, 2025

    Zongyu Lin, Yao Tang, Xingcheng Yao, Da Yin, Ziniu Hu, Yizhou Sun, and Kai-Wei Chang. Qlass: Boosting language agent inference via q-guided stepwise search.arXiv preprint arXiv:2502.02584, 2025

    Pith/arXiv arXiv 2025

  68. [68]

    Costbench: Evaluating multi-turn cost-optimal planning and adaptation in dynamic environments for llm tool-use agents.arXiv preprint arXiv:2511.02734, 2025

    Jiayu Liu, Cheng Qian, Zhaochen Su, Qing Zong, Shijue Huang, Bingxiang He, and Yi R Fung. Costbench: Evaluating multi-turn cost-optimal planning and adaptation in dynamic environments for llm tool-use agents.arXiv preprint arXiv:2511.02734, 2025

    Pith/arXiv arXiv 2025

  69. [69]

    Rcr-router: Efficient role-aware context routing for multi-agent llm systems with structured memory.arXiv preprint arXiv:2508.04903, 2025

    Jun Liu, Zhenglun Kong, Changdi Yang, Fan Yang, Tianqi Li, Peiyan Dong, Joannah Nanjekye, Hao Tang, Geng Yuan, Wei Niu, et al. Rcr-router: Efficient role-aware context routing for multi-agent llm systems with structured memory.arXiv preprint arXiv:2508.04903, 2025

    Pith/arXiv arXiv 2025

  70. [70]

    Toolscope: Enhancing llm agent tool use through tool merging and context-aware filtering,

    Marianne Menglin Liu, Daniel Garcia, Fjona Parllaku, Vikas Upadhyay, Syed Fahad Allam Shah, and Dan Roth. Toolscope: Enhancing llm agent tool use through tool merging and context-aware filtering,

  71. [71]

    Lost in the middle: How language models use long contexts.Transactions of the Association for Computational Linguistics, 12:157–173, 2024

    Nelson F Liu, Kevin Lin, John Hewitt, Ashwin Paranjape, Michele Bevilacqua, Fabio Petroni, and Percy Liang. Lost in the middle: How language models use long contexts.Transactions of the Association for Computational Linguistics, 12:157–173, 2024

    2024

  72. [72]

    Budget-aware tool-use enables effective agent scaling, 2025

    Tengxiao Liu, Zifeng Wang, Jin Miao, I-Hung Hsu, Jun Yan, Jiefeng Chen, Rujun Han, Fangyuan Xu, Yanfei Chen, Ke Jiang, Samira Daruki, Yi Liang, William Yang Wang, Tomas Pfister, and Chen-Yu Lee. Budget-aware tool-use enables effective agent scaling, 2025. URLhttps://arxiv.org/abs/ 2511.17006

    arXiv 2025

  73. [73]

    Groupdebate: Enhancing the efficiency of multi-agent debate using group discussion.arXiv preprint arXiv:2409.14051, 2024

    Tongxuan Liu, Xingyu Wang, Weizhe Huang, Wenjiang Xu, Yuting Zeng, Lei Jiang, Hailong Yang, and Jing Li. Groupdebate: Enhancing the efficiency of multi-agent debate using group discussion.arXiv preprint arXiv:2409.14051, 2024

    arXiv 2024

  74. [74]

    Infiguiagent: A multimodal generalist gui agent with native reasoning and reflection, 2025

    Yuhang Liu, Pengxiang Li, Zishu Wei, Congkai Xie, Xueyu Hu, Xinchen Xu, Shengyu Zhang, Xiaotian Han, Hongxia Yang, and Fei Wu. Infiguiagent: A multimodal generalist gui agent with native reasoning and reflection, 2025. URLhttps://arxiv.org/abs/2501.04575

    Pith/arXiv arXiv 2025

  75. [75]

    The ai scientist: Towards fully automated open-ended scientific discovery.arXiv preprint arXiv:2408.06292, 2024

    Chris Lu, Cong Lu, Robert Tjarko Lange, Jakob Foerster, Jeff Clune, and David Ha. The ai scientist: Towards fully automated open-ended scientific discovery.arXiv preprint arXiv:2408.06292, 2024

    Pith/arXiv arXiv 2024

  76. [76]

    Memochat: Tuning llms to use memos for consistent long-range open-domain conversation.arXiv preprint arXiv:2308.08239, 2023

    Junru Lu, Siyu An, Mingbao Lin, Gabriele Pergola, Yulan He, Di Yin, Xing Sun, and Yunsheng Wu. Memochat: Tuning llms to use memos for consistent long-range open-domain conversation.arXiv preprint arXiv:2308.08239, 2023. 41

    Pith/arXiv arXiv 2023

  77. [77]

    Toolshed: Scale tool-equipped agents with advanced rag-tool fusion and tool knowledge bases

    Elias Lumer, Vamse Kumar Subbiah, James A Burke, Pradeep Honaganahalli Basavaraju, and Austin Huber. Toolshed: Scale tool-equipped agents with advanced rag-tool fusion and tool knowledge bases. arXiv preprint arXiv:2410.14594, 2024

    Pith/arXiv arXiv 2024

  78. [78]

    Sciagent: Tool-augmented language models for scientific reasoning.arXiv preprint arXiv:2402.11451, 2024

    Yubo Ma, Zhibin Gou, Junheng Hao, Ruochen Xu, Shuohang Wang, Liangming Pan, Yujiu Yang, Yixin Cao, Aixin Sun, Hany Awadalla, et al. Sciagent: Tool-augmented language models for scientific reasoning.arXiv preprint arXiv:2402.11451, 2024

    Pith/arXiv arXiv 2024

  79. [79]

    Evaluating very long-term conversational memory of LLM agents

    Adyasha Maharana, Dong-Ho Lee, Sergey Tulyakov, Mohit Bansal, Francesco Barbieri, and Yuwei Fang. Evaluating very long-term conversational memory of LLM agents. In Lun-Wei Ku, Andre Martins, and Vivek Srikumar, editors,Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 13851–13870, Bangko...

    doi:10.18653/v1/2024.acl-long.747 2024

  80. [80]

    Gaia: a benchmarkforgeneralaiassistants

    Grégoire Mialon, Clémentine Fourrier, Thomas Wolf, Yann LeCun, and Thomas Scialom. Gaia: a benchmarkforgeneralaiassistants. InTheTwelfthInternationalConferenceonLearningRepresentations, 2023

    2023

Showing first 80 references.