Pith. sign in

REVIEW 5 major objections 5 minor 194 references

CommitKV: Lifecycle-Aware KV Cache Compression via Commit Transitions for Multi-Turn Agents

T0 review · 5 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read A cache-compression method that retires tool-call pages only after their observed role ends keeps agent accuracy near full-cache levels.

desk verdict The commit-transition idea is new and worth engaging, but the paper is not reproducible yet because the fixed-budget selection step is undefined and the empirical claims lack error bars. read the letter →

arxiv 2608.07855 v1 pith:UDOSWBRS submitted 2026-08-08 cs.LG

Weizhong Huang , Jinchao Zhang , Xiawu Zheng This is my paper · ORCID
classification cs.LG
keywords KVcachecompressionmulti-turnagentsReActlifecycle-awareevictioncommittransitionsdeletioneffecttoolcallingLLMinference
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

CommitKV is a method for compressing the key-value cache of multi-turn ReAct agents that decides what to evict from a page's lifecycle, not from its current attention score. It measures how much the attention output would change if a page of tool-call or observation tokens were deleted, once just before the tool-call commit and once right after the returned observation is processed. A page whose deletion effect drops from high to low across that boundary is treated as having completed its role and becomes a candidate for retirement, while a low-to-low page is treated as dormant and kept. Retirement candidates are only removed after a joint validation bounds the cumulative effect of deleting them together. The paper reports that this keeps agent accuracy close to the uncompressed cache while cutting peak KV memory by up to 5.00x and speeding up end-to-end inference by up to 5.62x.

What carries the argument

The commit transition is the central mechanism: for each tool call, the deletion effect of the same event page is measured in the pre-commit window (last W decode queries of the tool call) and the post-commit window (first W next-turn queries after the observation is appended). Deletion effect R(E;Q) is the maximum, over query positions and attention heads, of the relative change in an attention head's output when the page's tokens are deleted and remaining attention weights renormalized. Comparing the paired measurements through absolute thresholds and percentile ranks yields the lifecycle state that separates dormant from completed pages.

What would settle it

Run a trajectory where a tool-call page has low deletion effect over the first 8 post-commit queries but is needed by query 9 or in a later turn, for example the model re-reads the tool arguments after the observation; if CommitKV retires that page and the final answer's accuracy drops below FullKV while a variant that protects it does not, the lifecycle signal is not sufficient.

Watch

Extended reading notes

Core claim

The central claim is that snapshot-based importance scores conflate two situations that lifecycle signals separate: an event page may be temporarily dormant but needed later, or it may have finished its observed role after the tool's observation arrives. CommitKV establishes this separation by computing a deletion effect R(E;Q) for each eligible page over paired query windows on either side of a tool-call commit, combining absolute effect with percentile rank so both large-but-common and small-but-rare effects are labeled consistently. The lifecycle state lambda_c(E) classifies the page as completion candidate (high before, low after), dormant (low before, low after), newly active, still active, or uncertain, and only completion candidates proceed to a greedy joint test. A page is retired only if its own post-commit effect and the cumulative effect of already accepted pages stay below a threshold, and pages awaiting post-commit measurement are protected within a reserved budget. The authors argue this makes eviction decisions depend on whether information has completed its observed role rather than on whether it looks important right now.

Load-bearing premise

The method assumes that a page whose deletion effect stays low during the first W=8 next-turn queries after the returned observation has finished its role, so a page that only becomes important later would be wrongly retired.

Editorial extensions

If this is right

  • With lifecycle-aware retirement, agent accuracy stays close to FullKV while using a fixed token budget, because dormant pages are kept and only role-completed pages are removed.
  • Snapshot-based baselines lose accuracy on tool-heavy benchmarks because they cannot distinguish dormant from completed pages; CommitKV's gap over the strongest baseline is up to 22.24 percentage points.
  • Retiring pages only after joint validation and protecting pending pages keeps the cache within budget without sacrificing the information needed to make future lifecycle decisions.
  • The method is robust across cache budgets of 2048, 4096, and 8192 tokens and across three tested backbones, with the largest gains at the most aggressive budget.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A possible extension would make the post-commit window adaptive, re-measuring pages that are still low after W=8 queries at later checkpoints, so pages that become important later are not retired prematurely.
  • The deletion effect is computed on one attention layer with unspecified layer choice; averaging the high-to-low signal across layers or selecting the layer that best predicts held-out accuracy would likely make lifecycle labels more robust.
  • The commit-transition idea could generalize beyond tool calls to any agent or program state with an explicit commit boundary, such as code execution or document editing, where earlier context has served its purpose.
  • A testable prediction is that the accuracy benefit of CommitKV over snapshot baselines grows with the number of tool calls per trajectory and shrinks on single-turn reasoning tasks, consistent with the paper's tool/evidence versus reasoning benchmark split.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 5 minor

Summary. The paper introduces CommitKV, a lifecycle-aware KV cache compression method for multi-turn ReAct agents. It partitions completed tool calls and returned observations into event pages, measures each page's deletion effect in a pre-commit query window (the last W decoding queries of the tool call) and a post-commit window (the first W queries of the next turn), and uses the paired high-to-low transition as evidence that the page's role is complete. Completion candidates are then accepted into a retirement set via a greedy joint-validation test that bounds the combined post-commit deletion effect; pages awaiting post-commit measurement are temporarily protected within a reserved budget. At compression checkpoints, retired pages are excluded, protected pages are retained, and the remaining resident tokens are selected under a token budget, with the same index set applied to keys, values, and absolute positions. The method is evaluated on three main LLM backbones (with three additional backbones reported in the appendix) across eight benchmarks and cache budgets of 2048, 4096, and 8192 tokens, reporting accuracy, peak KV memory, and end-to-end latency, along with ablations against five reduced variants.

Significance. If the claims hold, CommitKV addresses a real limitation of snapshot-based KV eviction for agentic workloads: low current attention does not distinguish temporarily dormant information from information whose role has finished. The page-level paired deletion-effect measurement and the joint validation step are a sensible and non-obvious design, and the paper gives credit to a concrete mechanism rather than an abstract intuition. The ablations in Table 3 provide initial evidence that each component contributes to the reported accuracy. The main-table results show large gains over strong baselines under aggressive budgets, and the memory/latency figures indicate practical benefit. However, the paper's central scientific claim—that CommitKV can distinguish dormant from completed information—is stronger than what the current measurements establish, and the missing specification of the fixed-budget selection rule makes the method incompletely defined. The lack of statistical uncertainty, code release, validation-set discipline for the many thresholds, and sensitivity analysis further limits the strength of the empirical evidence.

major comments (5)
  1. [§3.3, Eq. (13) and Algorithm 1, line 22] The selection of the retained-token index set I_j is never specified. Eq. (13) only states the constraints N_j ⊆ I_j and |I_j| ≤ B, and Algorithm 1 line 22 says to 'select I_j' without defining the ordering, scoring, or tie-breaking rule for the resident tokens that are neither retired nor protected. This is load-bearing because the actual cache contents, and therefore the accuracy and memory results, depend entirely on how the remaining budget is filled. Without this rule, the method is not reproducible and cannot be compared fairly with the baselines, which do specify their selection criteria. Please provide the exact selection algorithm, and clarify whether that selection is part of CommitKV or an inherited baseline mechanism.
  2. [§3.2, Eqs. (6)–(7)] The deletion effect R(E; Q) is computed from 'one attention layer', but the paper never identifies which layer is used. Attention distributions differ sharply across layers, so the high/low labels in Eq. (9) and the resulting retirement decisions may be artifacts of an arbitrary layer choice. Please state the layer selection rule and justify it, ideally with an ablation across layers or with an aggregation over layers that is invariant to the choice.
  3. [§3.1, Eq. (5) and §3.3, Eq. (11)] The post-commit window Q_c^+ contains only the first W=8 decoding queries after the returned observation is appended. The joint validation in Eq. (11) bounds only the combined post-commit deletion effect over that same window. A page whose role resumes after the first eight tokens of the next turn, or in a later turn, will have a low measured R_c^+(E) and, if its pre-commit effect was high, will be classified as a completion candidate and retired. Thus the measured high-to-low transition does not establish that the page has 'completed its observed role'; it establishes only that the page is not attended in the first eight post-commit queries. The abstract's central claim is therefore stronger than what the current measurement supports. Please either temper the claim or add direct evidence that pages retired by this rule are not needed later, for example by measuring the deletion effect over longer post-commit windows or by tracking whether retired pages would have been attended in subsequent turns.
  4. [§4.1 Implementation Details] The method introduces at least six thresholds (τ_use, τ_dead, τ_joint, ρ_use, ρ_dead, η) plus W, G, the scan cap, and the pending cap, and the paper fixes them without any validation split or sensitivity analysis. The reported accuracy gains could in part encode tuning to the test benchmarks. Please report how these values were chosen, add a sensitivity study over the most influential thresholds, and, if possible, evaluate on a held-out validation set or report cross-benchmark stability.
  5. [§4.1–§4.4, Tables 1–3 and Figure 3] All accuracy, memory, and latency numbers are reported as single point estimates without error bars, repeated seeds, or significance tests. Given the small per-dataset sample sizes and the fact that several reported differences are only a few percentage points, this makes it hard to judge whether CommitKV's advantage over the strongest baselines is reliable. Please provide variance estimates or confidence intervals, and release the evaluation code and prompts to make the results reproducible.
minor comments (5)
  1. [Table 1] Several entries in the Qwen3.6-27B block are malformed, e.g., '74.1122.0029.29' in the R-KV row and 'R-KV 10.924.8562.69' in the DeepSeek row; these need to be split into separate percentages.
  2. [Algorithm 1, line 5] The line says to compute R_ct-1^+(E) and P_ct-1^±(E), but P^- was already computed at the commit; please clarify which quantities are newly computed at this step and which are reused.
  3. [§4.1] The text repeatedly refers to an 'Appendix' for complete protocol and additional backbones, but the submission as provided does not include that appendix; please ensure the appendix is present in the final version.
  4. [Figure 3] The bar chart reports single-run wall-clock times; adding error bars or a note on measurement variance would make the speedup claims more credible.
  5. [Abstract and §5] The phrase 'ensure that CommitKV can distinguish dormant information from information that has completed its observed role' is too strong given the W=8 window and the missing fixed-budget selection rule; a more cautious formulation would better match the evidence.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the lifecycle labels are operational definitions, while the safety claim is evaluated against external benchmarks and baselines.

full rationale

CommitKV's derivation chain is self-contained: Section 3.1 defines the paired query windows (Eq. 5), Section 3.2 defines the deletion effect (Eqs. 6-7) and lifecycle states (Eqs. 8-10), and Section 3.3 defines the joint validation and protection mechanisms (Eqs. 11-13). None of these equations is fitted to the reported benchmarks or defined in terms of the accuracy outcomes they are used to predict; 'completion candidate' is an operational label derived from pre/post deletion effects, and the claim that such pages are safe to remove is tested against FullKV and external baselines on FRAMES, GAIA, ToolHop, xbench-DeepSearch, GPQA, Bamboogle, MATH-500, and AIME25. The fixed hyperparameters (W=8, G=16, eta=0.125, tau thresholds) are chosen without a reported validation split, which is a tuning and robustness concern rather than a circularity. The W=8 post-commit window and the unspecified attention layer are empirical assumptions about temporal locality, not self-referential reductions. The only self-citations appear in the related-work survey and are not load-bearing for the method or its evaluation. No step reduces to its own input by construction.

Assumptions & free parameters 10 free parameters · 4 assumptions · 0 invented entities

The central claim rests on the domain assumption that deletion-effect transitions track task relevance, plus six or more hand-set thresholds and several capacity caps. None of these are derived from first principles or tested with sensitivity analysis. No new physical entities are introduced; event page, commit transition, and lifecycle state are operational definitions within the algorithm.

free parameters (10)
  • tau_use = 0.05
    Absolute deletion-effect threshold for HIGH state; set in Section 4.1 with no sensitivity analysis.
  • tau_dead = 0.01
    Absolute deletion-effect threshold for LOW state; same section.
  • tau_joint = 0.01
    Joint-validation threshold in Eq. (11); set to the same fixed value.
  • rho_use = 0.75
    Percentile-rank threshold for HIGH state; Section 4.1.
  • rho_dead = 0.25
    Percentile-rank threshold for LOW state; Section 4.1.
  • eta = 0.125
    Fraction of budget reserved for pending pages, Eq. (12).
  • W = 8
    Query-window size in tokens for pre/post-commit measurement, Eq. (5).
  • G = 16
    Maximum event-page size in tokens; Section 4.1.
  • scan and pending caps = 64 scanned pages, 16 protected pending pages
    Implementation caps at each commit; Section 4.1.
  • attention layer for deletion effect = unspecified
    Section 3.2 measures R(E;Q) using one attention layer but never says which layer or how it was chosen.
assumptions (4)
  • domain assumption A high-to-low change in deletion effect across a commit indicates the page has completed its observed role (Eq. 10).
    This is the lifecycle hypothesis; it is asserted in Section 3.2 and illustrated in Figure 1, not derived or validated against ground-truth page roles.
  • domain assumption The first W=8 next-turn queries after the returned observation capture the post-commit influence of a page.
    Eq. (5) defines Q_c^+ as the first W decoding queries; a page that matters only later will be mislabeled as low.
  • domain assumption Max relative attention-output change on one layer and head is a valid proxy for page importance.
    Eqs. (6) and (7) take the max over heads but use only one attention layer; no layer is specified.
  • standard math Attention renormalization after deleting a page follows Eq. (6).
    The formula is exact if remaining attention weights scale proportionally, but it ignores possible shifts in softmax behavior or attention sinks; it is a standard approximation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of CommitKV: Lifecycle-Aware KV Cache Compression via Commit Transitions for Multi-Turn Agents." pith.science (2026). https://pith.science/paper/UDOSWBRS

@misc{pith2026260807855,
  author       = {Pith},
  title        = {Pith review of: CommitKV: Lifecycle-Aware KV Cache Compression via Commit Transitions for Multi-Turn Agents},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UDOSWBRS}},
  note         = {Machine review of arXiv:2608.07855}
}
read the original abstract

Multi-turn Reasoning-and-Acting (ReAct) agents accumulate growing trajectories of reasoning, tool calls, and observations. Their key-value (KV) caches grow accordingly, increasing memory use and attention cost during model inference. Existing KV cache compression methods reduce these costs by evicting states with low attention scores. However, low attention in the current turn does not imply future irrelevance, as temporarily inactive information may become important later. Snapshot-based eviction methods therefore do not explicitly distinguish temporarily dormant information from information that appears to have completed its role. In this paper, we present CommitKV, which identifies KV lifecycles through commit transitions. Specifically, CommitKV first divides completed agent events into token pages and compares each eligible page's deletion effect before a tool-call commit and after the commit's returned observation has been incorporated. Based on these paired measurements, CommitKV distinguishes dormant pages from high-to-low completion candidates. It then applies a greedy joint test, accepting candidates for retirement only when their combined post-commit effect remains bounded. Finally, at a later compression checkpoint, accepted pages are excluded, a bounded set of pages awaiting post-commit measurement is protected, and the remaining KV states are retained within the cache budget using the same token indices for keys, values, and absolute positions. These mechanisms ensure that CommitKV can distinguish dormant information from information that has completed its observed role and can be safely removed. Experiments on various benchmarks show that CommitKV reduces agent memory use, accelerates end-to-end inference, and achieves higher accuracy than existing KV cache compression methods.

Figures

Figures reproduced from arXiv: 2608.07855 by the authors.

Figure 1
Figure 1. Why snapshot scores are insufficient for KV-cache compression. At checkpoint [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Overview of CommitKV. (1) For each tool-call commit, the deletion effect of the same eligible event page is measured in the pre- and post-commit query windows. (2) The paired deletion effects and percentile ranks determine the page’s lifecycle state; only a high-to-low page becomes a completion candidate. (3) A greedy joint test constructs the retirement set Dc. At checkpoint j, tokens in Dc are excluded, pending in… view at source ↗
Figure 3
Figure 3. Peak per-GPU KV-cache memory and wall-clock time per completed sample for Qwen3- [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

194 extracted references · 16 canonical work pages

  1. [1]

    arXiv preprint arXiv:2603.16859 , year=

    Socialomni: Benchmarking audio-visual social interactivity in omni models , author=. arXiv preprint arXiv:2603.16859 , year=

    arXiv

  2. [2]

    arXiv preprint arXiv:2603.14807 , year=

    Himemvln: Enhancing reliability of open-source zero-shot vision-and-language navigation with hierarchical memory system , author=. arXiv preprint arXiv:2603.14807 , year=

    arXiv

  3. [3]

    Communication, Simulation, and Intelligent Agents: Implications of Personal Intelligent Machines for Medical Education

    Clancey, William J. Communication, Simulation, and Intelligent Agents: Implications of Personal Intelligent Machines for Medical Education. Proceedings of the Eighth International Joint Conference on Artificial Intelligence (IJCAI-83)

  4. [4]

    Classification Problem Solving

    Clancey, William J. Classification Problem Solving. Proceedings of the Fourth National Conference on Artificial Intelligence

  5. [5]

    , title =

    Robinson, Arthur L. , title =. 1980 , doi =. https://science.sciencemag.org/content/208/4447/1019.full.pdf , journal =

    1980

  6. [6]

    New Ways to Make Microcircuits Smaller---Duplicate Entry

    Robinson, Arthur L. New Ways to Make Microcircuits Smaller---Duplicate Entry. Science

  7. [7]

    Clancey and Glenn Rennels , abstract =

    Diane Warner Hasling and William J. Clancey and Glenn Rennels , abstract =. Strategic explanations for a diagnostic consultation system , journal =. 1984 , issn =. doi:https://doi.org/10.1016/S0020-7373(84)80003-6 , url =

    doi:10.1016/s0020-7373(84)80003-6 1984

  8. [8]

    and Rennels, Glenn R

    Hasling, Diane Warner and Clancey, William J. and Rennels, Glenn R. and Test, Thomas. Strategic Explanations in Consultation---Duplicate. The International Journal of Man-Machine Studies

Show all 194 references
  1. [9]

    Poligon: A System for Parallel Problem Solving

    Rice, James. Poligon: A System for Parallel Problem Solving

  2. [10]

    Transfer of Rule-Based Expertise through a Tutorial Dialogue

    Clancey, William J. Transfer of Rule-Based Expertise through a Tutorial Dialogue

  3. [11]

    The Engineering of Qualitative Models

    Clancey, William J. The Engineering of Qualitative Models

  4. [12]

    2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pages=

    Solving oscillation problem in post-training quantization through a theoretical perspective , author=. 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pages=. 2023 , organization=

    2023

  5. [13]

    2023 , eprint=

    Attention Is All You Need , author=. 2023 , eprint=

    2023

  6. [14]

    Pluto: The 'Other' Red Planet

    NASA. Pluto: The 'Other' Red Planet

  7. [15]

    Advances in neural information processing systems , volume=

    Attention is all you need , author=. Advances in neural information processing systems , volume=

  8. [16]

    arXiv preprint arXiv:2602.10825 , year=

    Flow caching for autoregressive video generation , author=. arXiv preprint arXiv:2602.10825 , year=

  9. [17]

    Advances in neural information processing systems , volume=

    Chain-of-thought prompting elicits reasoning in large language models , author=. Advances in neural information processing systems , volume=

  10. [18]

    International Conference on Learning Representations , volume=

    Affinequant: Affine transformation quantization for large language models , author=. International Conference on Learning Representations , volume=

  11. [19]

    The Eleventh International Conference on Learning Representations , year=

    Self-Consistency Improves Chain of Thought Reasoning in Language Models , author=. The Eleventh International Conference on Learning Representations , year=

  12. [20]

    The Eleventh International Conference on Learning Representations , year=

    Least-to-Most Prompting Enables Complex Reasoning in Large Language Models , author=. The Eleventh International Conference on Learning Representations , year=

  13. [21]

    The Eleventh International Conference on Learning Representations , year=

    ReAct: Synergizing Reasoning and Acting in Language Models , author=. The Eleventh International Conference on Learning Representations , year=

  14. [22]

    Advances in neural information processing systems , volume=

    Tree of thoughts: Deliberate problem solving with large language models , author=. Advances in neural information processing systems , volume=

  15. [23]

    Advances in neural information processing systems , volume=

    Toolformer: Language models can teach themselves to use tools , author=. Advances in neural information processing systems , volume=

  16. [24]

    arXiv preprint arXiv:2205.00445 , year=

    MRKL Systems: A modular, neuro-symbolic architecture that combines large language models, external knowledge sources and discrete reasoning , author=. arXiv preprint arXiv:2205.00445 , year=

    arXiv

  17. [25]

    arXiv preprint arXiv:2510.24701 , year=

    Tongyi deepresearch technical report , author=. arXiv preprint arXiv:2510.24701 , year=

    arXiv

  18. [26]

    arXiv preprint arXiv:2510.16079 , year=

    Evolver: Self-evolving llm agents through an experience-driven lifecycle , author=. arXiv preprint arXiv:2510.16079 , year=

    arXiv

  19. [27]

    arXiv preprint arXiv:2506.18096 , year=

    Deep research agents: A systematic examination and roadmap , author=. arXiv preprint arXiv:2506.18096 , year=

    arXiv

  20. [28]

    Advances in Neural Information Processing Systems , volume=

    Webthinker: Empowering large reasoning models with deep research capability , author=. Advances in Neural Information Processing Systems , volume=

  21. [29]

    A Survey on Large Language Model Acceleration based on

    Haoyang LI and Yiming Li and Anxin Tian and Tianhao Tang and Zhanchao Xu and Xuejia Chen and Nicole HU and Wei Dong and Li Qing and Lei Chen , journal=. A Survey on Large Language Model Acceleration based on. 2025 , url=

    2025

  22. [30]

    Frontiers of Computer Science , volume=

    A survey on large language model based autonomous agents , author=. Frontiers of Computer Science , volume=. 2024 , publisher=

    2024

  23. [31]

    Proceedings of the AAAI conference on artificial intelligence , volume=

    Ompq: Orthogonal mixed precision quantization , author=. Proceedings of the AAAI conference on artificial intelligence , volume=

  24. [32]

    Advances in Neural Information Processing Systems , volume=

    Swe-agent: Agent-computer interfaces enable automated software engineering , author=. Advances in Neural Information Processing Systems , volume=

  25. [33]

    Proceedings of machine learning and systems , volume=

    Efficiently scaling transformer inference , author=. Proceedings of machine learning and systems , volume=

  26. [34]

    arXiv preprint arXiv:2508.10016 , year=

    Training-Free Multimodal Large Language Model Orchestration , author=. arXiv preprint arXiv:2508.10016 , year=

    arXiv

  27. [35]

    International Conference on Machine Learning , pages=

    WorkArena: How Capable are Web Agents at Solving Common Knowledge Work Tasks? , author=. International Conference on Machine Learning , pages=. 2024 , organization=

    2024

  28. [36]

    Advances in Neural Information Processing Systems , volume=

    Mind2web: Towards a generalist agent for the web , author=. Advances in Neural Information Processing Systems , volume=

  29. [37]

    Advances in neural information processing systems , volume=

    Agentboard: An analytical evaluation board of multi-turn llm agents , author=. Advances in neural information processing systems , volume=

  30. [38]

    Science China Information Sciences , volume=

    The rise and potential of large language model based agents: A survey , author=. Science China Information Sciences , volume=. 2025 , publisher=

    2025

  31. [39]

    Advances in neural information processing systems , volume=

    Reflexion: Language agents with verbal reinforcement learning , author=. Advances in neural information processing systems , volume=

  32. [40]

    The Third Conference on Parsimony and Learning (Proceedings Track) , year=

    Sparsity-Aware Prompt Tuning: A Simple and Effective Way to Fine-tune High-Sparsity LLMs , author=. The Third Conference on Parsimony and Learning (Proceedings Track) , year=

  33. [41]

    Advances in neural information processing systems , volume=

    Self-refine: Iterative refinement with self-feedback , author=. Advances in neural information processing systems , volume=

  34. [42]

    Transactions on Machine Learning Research , issn=

    Voyager: An Open-Ended Embodied Agent with Large Language Models , author=. Transactions on Machine Learning Research , issn=. 2024 , url=

    2024

  35. [43]

    arXiv preprint arXiv:2112.09332 , year=

    Webgpt: Browser-assisted question-answering with human feedback , author=. arXiv preprint arXiv:2112.09332 , year=

    arXiv

  36. [44]

    arXiv preprint arXiv:2310.08560 , year=

    MemGPT: Towards LLMs as Operating Systems , author=. arXiv preprint arXiv:2310.08560 , year=

    arXiv

  37. [45]

    International Conference on Learning Representations , volume=

    Agentbench: Evaluating llms as agents , author=. International Conference on Learning Representations , volume=

  38. [46]

    arXiv preprint arXiv:2603.08754 , year=

    Hindsight Credit Assignment for Long-Horizon LLM Agents , author=. arXiv preprint arXiv:2603.08754 , year=

  39. [47]

    arXiv preprint arXiv:2603.04257 , year=

    Memex (rl): Scaling long-horizon llm agents via indexed experience memory , author=. arXiv preprint arXiv:2603.04257 , year=

  40. [48]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    AutoTool: Efficient tool selection for large language model agents , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  41. [49]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    DesireKV: Decoupling Sensitivity and Importance for Reasoning-Aware KV Cache Compression , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  42. [50]

    arXiv preprint arXiv:2403.04643 , year=

    Qaq: Quality adaptive quantization for llm kv cache , author=. arXiv preprint arXiv:2403.04643 , year=

    arXiv

  43. [51]

    International Conference on Machine Learning , pages=

    KIVI: A Tuning-Free Asymmetric 2bit Quantization for KV Cache , author=. International Conference on Machine Learning , pages=. 2024 , organization=

    2024

  44. [52]

    arXiv preprint arXiv:2606.03099 , year=

    Photocraft: Agentic reasoning with hierarchical self-evolving memory for deep image search , author=. arXiv preprint arXiv:2606.03099 , year=

    arXiv

  45. [53]

    International Conference on Machine Learning , pages=

    LongRoPE: Extending LLM Context Window Beyond 2 Million Tokens , author=. International Conference on Machine Learning , pages=. 2024 , organization=

    2024

  46. [54]

    arXiv preprint arXiv:2509.06501 , year=

    Webexplorer: Explore and evolve for training long-horizon web agents , author=. arXiv preprint arXiv:2509.06501 , year=

  47. [55]

    arXiv preprint arXiv:2506.13585 , year=

    Minimax-m1: Scaling test-time compute efficiently with lightning attention , author=. arXiv preprint arXiv:2506.13585 , year=

    arXiv

  48. [56]

    ACM Computing Surveys , year=

    Scaling Up Multi-Agent Reinforcement Learning for Large Agent Teams and Long-Horizon Tasks: A Survey , author=. ACM Computing Surveys , year=

  49. [57]

    arXiv preprint arXiv:2502.01600 , year=

    Reinforcement learning for long-horizon interactive llm agents , author=. arXiv preprint arXiv:2502.01600 , year=

    arXiv

  50. [58]

    arXiv preprint arXiv:2603.00729 , year=

    Qwen3-coder-next technical report , author=. arXiv preprint arXiv:2603.00729 , year=

    arXiv

  51. [59]

    arXiv preprint arXiv:2402.03300 , year=

    Deepseekmath: Pushing the limits of mathematical reasoning in open language models , author=. arXiv preprint arXiv:2402.03300 , year=

    arXiv

  52. [60]

    arXiv preprint arXiv:2405.04434 , year=

    Deepseek-v2: A strong, economical, and efficient mixture-of-experts language model , author=. arXiv preprint arXiv:2405.04434 , year=

    arXiv

  53. [61]

    arXiv preprint arXiv:2302.13971 , year=

    Llama: Open and efficient foundation language models , author=. arXiv preprint arXiv:2302.13971 , year=

    arXiv

  54. [62]

    arXiv preprint arXiv:2605.09252 , year=

    LLM Agents Already Know When to Call Tools--Even Without Reasoning , author=. arXiv preprint arXiv:2605.09252 , year=

    arXiv

  55. [63]

    Advances in Neural Information Processing Systems , volume=

    Gorilla: Large language model connected with massive apis , author=. Advances in Neural Information Processing Systems , volume=

  56. [64]

    Advances in Neural Information Processing Systems , volume=

    H2o: Heavy-hitter oracle for efficient generative inference of large language models , author=. Advances in Neural Information Processing Systems , volume=

  57. [65]

    Advances in Neural Information Processing Systems , volume=

    Scissorhands: Exploiting the persistence of importance hypothesis for llm kv cache compression at test time , author=. Advances in Neural Information Processing Systems , volume=

  58. [66]

    Proceedings of the ACM on Management of Data , volume=

    Pqcache: Product quantization-based kvcache for long context llm inference , author=. Proceedings of the ACM on Management of Data , volume=. 2025 , publisher=

    2025

  59. [67]

    arXiv preprint arXiv:2407.18003 , year=

    Keep the cost down: A review on methods to optimize LLM's KV-cache consumption , author=. arXiv preprint arXiv:2407.18003 , year=

    arXiv

  60. [68]

    arXiv preprint arXiv:2309.16609 , year=

    Qwen technical report , author=. arXiv preprint arXiv:2309.16609 , year=

    arXiv

  61. [69]

    Forty-first International Conference on Machine Learning , year=

    Outlier-aware slicing for post-training quantization in vision transformer , author=. Forty-first International Conference on Machine Learning , year=

  62. [70]

    International Conference on Machine Learning , pages=

    QUEST: Query-Aware Sparsity for Efficient Long-Context LLM Inference , author=. International Conference on Machine Learning , pages=. 2024 , organization=

    2024

  63. [71]

    International Conference on Learning Representations , volume=

    Model tells you what to discard: Adaptive kv cache compression for llms , author=. International Conference on Learning Representations , volume=

  64. [72]

    Advances in Neural Information Processing Systems , volume=

    Chunkkv: Semantic-preserving kv cache compression for efficient long-context llm inference , author=. Advances in Neural Information Processing Systems , volume=

  65. [73]

    arXiv preprint arXiv:2401.03428 , year=

    Exploring large language model based intelligent agents: Definitions, methods, and prospects , author=. arXiv preprint arXiv:2401.03428 , year=

    arXiv

  66. [74]

    arXiv preprint arXiv:2509.12867 , year=

    Tool-R1: Sample-Efficient Reinforcement Learning for Agentic Tool Use , author=. arXiv preprint arXiv:2509.12867 , year=

  67. [75]

    arXiv preprint arXiv:2402.01680 , year=

    Large language model based multi-agents: A survey of progress and challenges , author=. arXiv preprint arXiv:2402.01680 , year=

    arXiv

  68. [76]

    Measuring and Narrowing the Compositionality Gap in Language Models

    Press, Ofir and Zhang, Muru and Min, Sewon and Schmidt, Ludwig and Smith, Noah and Lewis, Mike. Measuring and Narrowing the Compositionality Gap in Language Models. Findings of the Association for Computational Linguistics: EMNLP 2023. 2023. doi:10.18653/v1/2023.findings-emnlp.378

    2023 doi

  69. [77]

    The Fourteenth International Conference on Learning Representations , year=

    DefensiveKV: Taming the Fragility of KV Cache Eviction in LLM Inference , author=. The Fourteenth International Conference on Learning Representations , year=

  70. [78]

    The Fourteenth International Conference on Learning Representations , year=

    PM-KVQ: Progressive Mixed-precision KV Cache Quantization for Long-CoT LLMs , author=. The Fourteenth International Conference on Learning Representations , year=

  71. [79]

    The Fourteenth International Conference on Learning Representations , year=

    ReST-KV: Robust KV Cache Eviction with Layer-wise Output Reconstruction and Spatial-Temporal Smoothing , author=. The Fourteenth International Conference on Learning Representations , year=

  72. [80]

    arXiv preprint arXiv:2505.09388 , year=

    Qwen3 technical report , author=. arXiv preprint arXiv:2505.09388 , year=

    arXiv

  73. [81]

    arXiv preprint arXiv:2502.17419 , year=

    From system 1 to system 2: A survey of reasoning large language models , author=. arXiv preprint arXiv:2502.17419 , year=

    arXiv

  74. [82]

    Proceedings of the 31st International Conference on Computational Linguistics , pages=

    Beyond chain-of-thought: A survey of chain-of-x paradigms for llms , author=. Proceedings of the 31st International Conference on Computational Linguistics , pages=

  75. [83]

    Nature , volume=

    DeepSeek-R1 incentivizes reasoning in LLMs through reinforcement learning , author=. Nature , volume=. 2025 , publisher=

    2025

  76. [84]

    Advances in Neural Information Processing Systems , volume=

    Minicache: Kv cache compression in depth dimension for large language models , author=. Advances in Neural Information Processing Systems , volume=

  77. [85]

    International Conference on Machine Learning , pages=

    CaM: Cache Merging for Memory-efficient LLMs Inference , author=. International Conference on Machine Learning , pages=. 2024 , organization=

    2024

  78. [86]

    International Conference on Learning Representations , volume=

    Palu: KV-cache compression with low-rank projection , author=. International Conference on Learning Representations , volume=

  79. [87]

    International Conference on Machine Learning , pages=

    ShadowKV: KV Cache in Shadows for High-Throughput Long-Context LLM Inference , author=. International Conference on Machine Learning , pages=. 2025 , organization=

    2025

  80. [88]

    Advances in Neural Information Processing Systems , volume=

    Loki: Low-rank keys for efficient sparse attention , author=. Advances in Neural Information Processing Systems , volume=

  81. [89]

    arXiv preprint arXiv:2412.19442 , year=

    A survey on large language model acceleration based on kv cache management , author=. arXiv preprint arXiv:2412.19442 , year=

    arXiv

  82. [90]

    International Conference on Learning Representations , volume=

    Efficient streaming language models with attention sinks , author=. International Conference on Learning Representations , volume=

  83. [91]

    Advances in Neural Information Processing Systems , volume=

    Snapkv: Llm knows what you are looking for before generation , author=. Advances in Neural Information Processing Systems , volume=

  84. [92]

    Zefan Cai and Yichi Zhang and Bofei Gao and Yuliang Liu and Yucheng Li and Tianyu Liu and Keming Lu and Wayne Xiong and Yue Dong and Junjie Hu and Wen Xiao , booktitle=. Pyramid. 2025 , url=

    2025

  85. [93]

    Advances in Neural Information Processing Systems , volume=

    Ada-kv: Optimizing kv cache eviction by adaptive budget allocation for efficient llm inference , author=. Advances in Neural Information Processing Systems , volume=

  86. [94]

    Advances in Neural Information Processing Systems , volume=

    Accurate KV cache eviction via anchor direction projection for efficient LLM inference , author=. Advances in Neural Information Processing Systems , volume=

  87. [95]

    Advances in Neural Information Processing Systems , volume=

    Keydiff: Key similarity-based kv cache eviction for long-context llm inference in resource-constrained environments , author=. Advances in Neural Information Processing Systems , volume=

  88. [96]

    arXiv preprint arXiv:2606.03928 , year=

    Value-Aware Stochastic KV Cache Eviction for Reasoning Models , author=. arXiv preprint arXiv:2606.03928 , year=

    arXiv

  89. [97]

    International Conference on Learning Representations , volume=

    Mint: Evaluating llms in multi-turn interaction with tools and language feedback , author=. International Conference on Learning Representations , volume=

  90. [98]

    International Conference on Learning Representations , volume=

    MetaGPT: Meta programming for a multi-agent collaborative framework , author=. International Conference on Learning Representations , volume=

  91. [99]

    Advances in neural information processing systems , volume=

    Infllm: Training-free long-context extrapolation for llms with an efficient context memory , author=. Advances in neural information processing systems , volume=

  92. [100]

    Advances in Neural Information Processing Systems , volume =

    R-KV: Redundancy-aware KV Cache Compression for Reasoning Models , author =. Advances in Neural Information Processing Systems , volume =

  93. [101]

    Advances in Neural Information Processing Systems , volume=

    Zipcache: Accurate and efficient kv cache quantization with salient token identification , author=. Advances in Neural Information Processing Systems , volume=

  94. [102]

    Advances in Neural Information Processing Systems , volume=

    Kvquant: Towards 10 million context length llm inference with kv cache quantization , author=. Advances in Neural Information Processing Systems , volume=

  95. [103]

    arXiv preprint arXiv:2403.05527 , year=

    Gear: An efficient kv cache compression recipe for near-lossless generative inference of llm , author=. arXiv preprint arXiv:2403.05527 , year=

    arXiv

  96. [104]

    International Conference on Machine Learning , pages=

    Flexgen: High-throughput generative inference of large language models with a single gpu , author=. International Conference on Machine Learning , pages=. 2023 , organization=

    2023

  97. [105]

    Proceedings of the 29th symposium on operating systems principles , pages=

    Efficient memory management for large language model serving with pagedattention , author=. Proceedings of the 29th symposium on operating systems principles , pages=

  98. [106]

    Advances in neural information processing systems , volume=

    Sglang: Efficient execution of structured language model programs , author=. Advances in neural information processing systems , volume=

  99. [107]

    Advances in neural information processing systems , volume=

    Flashattention: Fast and memory-efficient exact attention with io-awareness , author=. Advances in neural information processing systems , volume=

  100. [108]

    International Conference on Learning Representations , volume=

    Flashattention-2: Faster attention with better parallelism and work partitioning , author=. International Conference on Learning Representations , volume=

  101. [109]

    International Conference on Machine Learning , pages=

    Fast inference from transformers via speculative decoding , author=. International Conference on Machine Learning , pages=. 2023 , organization=

    2023

  102. [110]

    International Conference on Machine Learning , pages=

    Medusa: Simple LLM Inference Acceleration Framework with Multiple Decoding Heads , author=. International Conference on Machine Learning , pages=. 2024 , organization=

    2024

  103. [111]

    Proceedings of the 57th annual meeting of the association for computational linguistics , pages=

    Transformer-xl: Attentive language models beyond a fixed-length context , author=. Proceedings of the 57th annual meeting of the association for computational linguistics , pages=

  104. [112]

    International Conference on Learning Representations , year=

    Compressive Transformers for Long-Range Sequence Modelling , author=. International Conference on Learning Representations , year=

  105. [113]

    International Conference on Learning Representations , year=

    Reformer: The Efficient Transformer , author=. International Conference on Learning Representations , year=

  106. [114]

    arXiv preprint arXiv:2004.05150 , year=

    Longformer: The long-document transformer , author=. arXiv preprint arXiv:2004.05150 , year=

    2004 arXiv

  107. [115]

    Advances in neural information processing systems , volume=

    Big bird: Transformers for longer sequences , author=. Advances in neural information processing systems , volume=

  108. [116]

    International Conference on Learning Representations , year=

    Rethinking Attention with Performers , author=. International Conference on Learning Representations , year=

  109. [117]

    arXiv preprint arXiv:2006.04768 , year=

    Linformer: Self-attention with linear complexity , author=. arXiv preprint arXiv:2006.04768 , year=

    2006 arXiv

  110. [118]

    International Conference on Learning Representations , year=

    Memorizing Transformers , author=. International Conference on Learning Representations , year=

  111. [119]

    arXiv preprint arXiv:2305.16300 , year=

    Landmark attention: Random-access infinite context length for transformers , author=. arXiv preprint arXiv:2305.16300 , year=

    arXiv

  112. [120]

    International Conference on Learning Representations , volume=

    Longlora: Efficient fine-tuning of long-context large language models , author=. International Conference on Learning Representations , volume=

  113. [121]

    International Conference on Learning Representations , volume=

    Yarn: Efficient context window extension of large language models , author=. International Conference on Learning Representations , volume=

  114. [122]

    Neurocomputing , volume=

    Roformer: Enhanced transformer with rotary position embedding , author=. Neurocomputing , volume=. 2024 , publisher=

    2024

  115. [123]

    International Conference on Learning Representations , year=

    Train Short, Test Long: Attention with Linear Biases Enables Input Length Extrapolation , author=. International Conference on Learning Representations , year=

  116. [124]

    Proceedings of the 62nd annual meeting of the association for computational linguistics (volume 1: Long papers) , pages=

    Longbench: A bilingual, multitask benchmark for long context understanding , author=. Proceedings of the 62nd annual meeting of the association for computational linguistics (volume 1: Long papers) , pages=

  117. [125]

    2024 , url=

    Cheng-Ping Hsieh and Simeng Sun and Samuel Kriman and Shantanu Acharya and Dima Rekesh and Fei Jia and Boris Ginsburg , booktitle=. 2024 , url=

    2024

  118. [126]

    Transactions of the association for computational linguistics , volume=

    Lost in the middle: How language models use long contexts , author=. Transactions of the association for computational linguistics , volume=

  119. [127]

    First Conference on Language Modeling , year=

    Mamba: Linear-Time Sequence Modeling with Selective State Spaces , author=. First Conference on Language Modeling , year=

  120. [128]

    Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing , pages=

    Gqa: Training generalized multi-query transformer models from multi-head checkpoints , author=. Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing , pages=

    2023

  121. [129]

    Fact, fetch, and reason: A unified evaluation of retrieval-augmented generation , author=. Proceedings 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=

    2025

  122. [130]

    arXiv preprint arXiv:2504.12516 , year=

    Browsecomp: A simple yet challenging benchmark for browsing agents , author=. arXiv preprint arXiv:2504.12516 , year=

    arXiv

  123. [131]

    arXiv preprint arXiv:2508.06600 , year=

    Browsecomp-plus: A more fair and transparent evaluation benchmark of deep-research agent , author=. arXiv preprint arXiv:2508.06600 , year=

    arXiv

  124. [132]

    arXiv preprint arXiv:2108.08532 , year=

    An information theory-inspired strategy for automatic network pruning , author=. arXiv preprint arXiv:2108.08532 , year=

    arXiv

  125. [133]

    International Conference on Learning Representations , volume=

    Gaia: a benchmark for general ai assistants , author=. International Conference on Learning Representations , volume=

  126. [134]

    International Conference on Learning Representations , volume=

    Webarena: A realistic web environment for building autonomous agents , author=. International Conference on Learning Representations , volume=

  127. [135]

    Advances in Neural Information Processing Systems , volume=

    Osworld: Benchmarking multimodal agents for open-ended tasks in real computer environments , author=. Advances in Neural Information Processing Systems , volume=

  128. [136]

    Findings of the Association for Computational Linguistics: ACL 2024 , pages=

    Stabletoolbench: Towards stable large-scale benchmarking on tool learning of large language models , author=. Findings of the Association for Computational Linguistics: ACL 2024 , pages=

    2024

  129. [137]

    International Conference on Learning Representations , volume=

    Swe-bench: Can language models resolve real-world github issues? , author=. International Conference on Learning Representations , volume=

  130. [138]

    arXiv preprint arXiv:2303.08774 , year=

    Gpt-4 technical report , author=. arXiv preprint arXiv:2303.08774 , year=

    arXiv

  131. [139]

    A structural probe for finding syntax in word representations , author=. Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers) , pages=

    2019

  132. [140]

    Proceedings of the 57th annual meeting of the association for computational linguistics , pages=

    BERT rediscovers the classical NLP pipeline , author=. Proceedings of the 57th annual meeting of the association for computational linguistics , pages=

  133. [141]

    Language models as knowledge bases? , author=. Proceedings of the 2019 conference on empirical methods in natural language processing and the 9th international joint conference on natural language processing (EMNLP-IJCNLP) , pages=

    2019

  134. [142]

    Transactions of the Association for Computational Linguistics , volume=

    Amnesic probing: Behavioral explanation with amnesic counterfactuals , author=. Transactions of the Association for Computational Linguistics , volume=

  135. [143]

    Advances in neural information processing systems , volume=

    Retrieval-augmented generation for knowledge-intensive nlp tasks , author=. Advances in neural information processing systems , volume=

  136. [144]

    Proceedings of the 2020 conference on empirical methods in natural language processing (EMNLP) , pages=

    Dense passage retrieval for open-domain question answering , author=. Proceedings of the 2020 conference on empirical methods in natural language processing (EMNLP) , pages=

    2020

  137. [145]

    arXiv preprint arXiv:2502.14770 , year=

    Determining layer-wise sparsity for large language models through a theoretical perspective , author=. arXiv preprint arXiv:2502.14770 , year=

    arXiv

  138. [146]

    arXiv preprint arXiv:2312.10997 , year=

    Retrieval-Augmented Generation for Large Language Models: A Survey , author=. arXiv preprint arXiv:2312.10997 , year=

    arXiv

  139. [147]

    Proceedings of the 16th conference of the european chapter of the association for computational linguistics: main volume , pages=

    Leveraging passage retrieval with generative models for open domain question answering , author=. Proceedings of the 16th conference of the european chapter of the association for computational linguistics: main volume , pages=

  140. [148]

    International conference on machine learning , pages=

    Retrieval augmented language model pre-training , author=. International conference on machine learning , pages=. 2020 , organization=

    2020

  141. [149]

    arXiv preprint arXiv:2405.08944 , year=

    Challenges in deploying long-context transformers: A theoretical peak performance analysis , author=. arXiv preprint arXiv:2405.08944 , year=

    arXiv

  142. [150]

    arXiv preprint arXiv:2412.03594 , year=

    Batchllm: Optimizing large batched llm inference with global prefix sharing and throughput-oriented token batching , author=. arXiv preprint arXiv:2412.03594 , year=

    arXiv

  143. [151]

    Proceedings of Machine Learning and Systems , volume=

    Fasttree: Optimizing attention kernel and runtime for tree-structured llm inference , author=. Proceedings of Machine Learning and Systems , volume=

  144. [152]

    arXiv preprint arXiv:2602.22603 , year=

    SideQuest: Model-Driven KV Cache Management for Long-Horizon Agentic Reasoning , author=. arXiv preprint arXiv:2602.22603 , year=

  145. [153]

    2025 , howpublished =

    2025

  146. [154]

    Measuring Mathematical Problem Solving With the

    Hendrycks, Dan and Burns, Collin and Kadavath, Saurav and Arora, Akul and Basart, Steven and Tang, Eric and Song, Dawn and Steinhardt, Jacob , journal =. Measuring Mathematical Problem Solving With the

  147. [155]

    Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025. doi:10.18653/v1/2025.acl-long.150

    2025 doi

  148. [156]

    Li, Minghao and Zhao, Yingxiu and Yu, Bowen and Song, Feifan and Li, Hangyu and Yu, Haiyang and Li, Zhoujun and Huang, Fei and Li, Yongbin , booktitle =

  149. [157]

    Qin, Yujia and Liang, Shihao and Ye, Yining and Zhu, Kunlun and Yan, Lan and Lu, Yaxi and Lin, Yankai and Cong, Xin and Tang, Xiangru and Qian, Bill and others , booktitle =

  150. [158]

    Gemma 4 12B Unified Model , year =

  151. [159]

    arXiv preprint arXiv:2504.21318 , year=

    Phi-4-reasoning technical report , author=. arXiv preprint arXiv:2504.21318 , year=

    arXiv

  152. [160]

    International Conference on Learning Representations , volume=

    Livecodebench: Holistic and contamination free evaluation of large language models for code , author=. International Conference on Learning Representations , volume=

  153. [161]

    2021 , url=

    Mohit Shridhar and Xingdi Yuan and Marc-Alexandre Cote and Yonatan Bisk and Adam Trischler and Matthew Hausknecht , booktitle=. 2021 , url=

    2021

  154. [162]

    2025 , publisher =

    American Invitational Mathematics Examination (AIME) 2025 , author =. 2025 , publisher =

    2025

  155. [163]

    MiniCPM5-1B , year =

  156. [164]

    arXiv preprint arXiv:2508.10925 , year=

    gpt-oss-120b & gpt-oss-20b model card , author=. arXiv preprint arXiv:2508.10925 , year=

    arXiv

  157. [165]

    arXiv preprint arXiv:2501.14249 , year=

    Humanity's last exam , author=. arXiv preprint arXiv:2501.14249 , year=

    arXiv

  158. [166]

    Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , pages=

    Webwalker: Benchmarking llms in web traversal , author=. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , pages=

  159. [167]

    First Conference on Language Modeling , year=

    GPQA: A Graduate-Level Google-Proof Q&A Benchmark , author=. First Conference on Language Modeling , year=

  160. [168]

    arXiv preprint arXiv:2504.19314 , year=

    Browsecomp-zh: Benchmarking web browsing ability of large language models in chinese , author=. arXiv preprint arXiv:2504.19314 , year=

    arXiv

  161. [169]

    arXiv preprint arXiv:2506.13651 , year=

    xbench: Tracking agents productivity scaling with profession-aligned real-world evaluations , author=. arXiv preprint arXiv:2506.13651 , year=

    arXiv

  162. [170]

    arXiv preprint arXiv:2503.09572 , year=

    Plan-and-act: Improving planning of agents for long-horizon tasks , author=. arXiv preprint arXiv:2503.09572 , year=

    arXiv

  163. [171]

    International Conference on Learning Representations , volume=

    Dynamic low-rank sparse adaptation for large language models , author=. International Conference on Learning Representations , volume=

  164. [172]

    arXiv preprint arXiv:2604.04921 , year=

    Triattention: Efficient long reasoning with trigonometric kv compression , author=. arXiv preprint arXiv:2604.04921 , year=

    arXiv

  165. [173]

    arXiv preprint arXiv:2602.15763 , year=

    Glm-5: from vibe coding to agentic engineering , author=. arXiv preprint arXiv:2602.15763 , year=

    arXiv

  166. [174]

    arXiv preprint arXiv:2412.19437 , year=

    Deepseek-v3 technical report , author=. arXiv preprint arXiv:2412.19437 , year=

    arXiv

  167. [175]

    arXiv preprint arXiv:2603.11445 , year=

    Verified multi-agent orchestration: A plan-execute-verify-replan framework for complex query resolution , author=. arXiv preprint arXiv:2603.11445 , year=

  168. [176]

    arXiv preprint arXiv:2507.02592 , year=

    Websailor: Navigating super-human reasoning for web agent , author=. arXiv preprint arXiv:2507.02592 , year=

    arXiv

  169. [177]

    Advances in Neural Information Processing Systems , volume=

    Webdancer: Towards autonomous information seeking agency , author=. Advances in Neural Information Processing Systems , volume=

  170. [178]

    arXiv preprint arXiv:2503.05592 , year=

    R1-searcher: Incentivizing the search capability in llms via reinforcement learning , author=. arXiv preprint arXiv:2503.05592 , year=

    arXiv

  171. [179]

    arXiv preprint arXiv:2602.14721 , year=

    Webworld: A large-scale world model for web agent training , author=. arXiv preprint arXiv:2602.14721 , year=

  172. [180]

    arXiv preprint arXiv:2604.08516 , year=

    Molmoweb: Open visual web agent and open data for the open web , author=. arXiv preprint arXiv:2604.08516 , year=

    arXiv

  173. [181]

    Taowen Tony Liu and Jeffrey T. H. Wong and Can Xiao and Bowen Yang and Hao Mark Chen and Yiren Zhao , booktitle=. Agent. 2026 , url=

    2026

  174. [182]

    Proceedings of Machine Learning and Systems , volume=

    Rethinking Key-Value Cache Compression Techniques for Large Language Model Serving , author=. Proceedings of Machine Learning and Systems , volume=

  175. [183]

    arXiv preprint arXiv:2607.10582 , year=

    MemDecay: Region-Aware KV Cache Eviction for Efficient LLM Agent Inference , author=. arXiv preprint arXiv:2607.10582 , year=

    arXiv

  176. [184]

    arXiv preprint arXiv:2606.09916 , year=

    IntentKV: Cross-Turn Intent-Aware KV Cache Pruning for Agent Inference , author=. arXiv preprint arXiv:2606.09916 , year=

    arXiv

  177. [185]

    Findings of the Association for Computational Linguistics: EACL 2026 , pages=

    Pagedeviction: Structured block-wise kv cache pruning for efficient large language model inference , author=. Findings of the Association for Computational Linguistics: EACL 2026 , pages=

    2026

  178. [186]

    arXiv preprint arXiv:2509.17396 , year=

    EpiCache: Episodic KV Cache Management for Long-Term Conversation on Resource-Constrained Environments , author=. arXiv preprint arXiv:2509.17396 , year=

    arXiv

  179. [187]

    arXiv preprint arXiv:2403.17297 , year=

    Internlm2 technical report , author=. arXiv preprint arXiv:2403.17297 , year=

    arXiv

  180. [188]

    Long Hu and Nan Jia and Rui Wang and Jiahui Li and Qingyi Yang and Yixue Hao and Xianzhi Li and Xiaofei Liao , year=

  181. [189]

    2025 , eprint=

    CAOTE: KV Cache Selection for LLMs via Attention Output Error-Based Token Eviction , author=. 2025 , eprint=

    2025

  182. [190]

    Cache You Later: Post-Compression

    Andrew Rusli and Shreyan Paliwal and Michael Jiao , booktitle=. Cache You Later: Post-Compression. 2026 , url=

    2026

  183. [191]

    Advances in Neural Information Processing Systems , volume=

    Arkvale: Efficient generative llm inference with recallable key-value eviction , author=. Advances in Neural Information Processing Systems , volume=

  184. [192]

    arXiv preprint arXiv:2607.00692 , year=

    Self-GC: Self-Governing Context for Long-Horizon LLM Agents , author=. arXiv preprint arXiv:2607.00692 , year=

    arXiv

  185. [193]

    arXiv preprint arXiv:2404.11584 , year=

    The landscape of emerging ai agent architectures for reasoning, planning, and tool calling: A survey , author=. arXiv preprint arXiv:2404.11584 , year=

    arXiv

  186. [194]

    2024 USENIX annual technical conference (USENIX ATC 24) , pages=

    \ Cost-Efficient \ large language model serving for multi-turn conversations with \ CachedAttention \ , author=. 2024 USENIX annual technical conference (USENIX ATC 24) , pages=

    2024

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.