Pith. sign in

REVIEW 2 major objections 7 minor 71 references

Fast, Slow, and Tool-augmented Thinking for LLMs: A Review

T0 review · 2 major / 7 minor · reviewed 2026-08-15 · deepseek-v4-flash

Pith's one-line read Fast, slow, and tool-augmented thinking are not three separate LLM reasoning modes but two independent choices: how deeply to reason and whether to draw on external knowledge.

desk verdict Useful two-axis survey of adaptive LLM reasoning, worth refereeing despite an over-stated orthogonality claim and thin positioning against prior surveys. read the letter →

arxiv 2508.12265 v2 pith:OPLFKFBY submitted 2025-08-17 cs.CL

classification cs.CL
keywords LLMreasoningadaptivefastandslowthinkingtool-augmentedchain-of-thoughtretrieval-augmentedgenerationstrategyselectiondual-processtheory
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

This review argues that the three familiar modes of LLM reasoning—fast, slow, and tool-augmented—are better described by two orthogonal boundaries than by three separate strategies. The fast/slow boundary separates immediate intuitive answers from deliberate step-by-step chains; the internal/external boundary separates knowledge stored in model weights from knowledge obtained through tools, retrieval, or computation. Tool-augmented thinking is not a third mode but an external-knowledge axis that can intersect with either depth. The paper also unifies the growing collection of adaptive-selection methods under one two-stage abstraction—extract decision factors, then compare scores against thresholds—and claims to be the first review covering adaptive selection among all three strategies. If the taxonomy holds, reasoning depth and knowledge source become independently tunable levers for routing queries.

What carries the argument

The central object is the pair of orthogonal knowledge boundaries that define the 2×2 strategy space: a fast/slow boundary separating intuitive direct responses from deliberative step-by-step reasoning, and an internal/external boundary separating parametric knowledge from tool-augmented knowledge. The argument is carried by the threshold-crossing decision functions of Eqs. (5)–(6): a decision score $\Phi_d$ or $\Phi_s$ is computed from extracted factors (confidence, task complexity, utility gain) and compared with a threshold $\delta_d$ or $\delta_s$; crossing the boundary flips the strategy on that axis. This mechanism is flexible enough to cover learned implicit policies, where the threshold lives inside the model weights, and explicit routers, where the threshold is a hand-set scalar or a trained lightweight classifier.

What would settle it

On the same set of mixed-difficulty, mixed-tool queries, measure whether the optimal fast/slow switching threshold shifts when retrieval is available versus withheld; a systematic shift would mean the two axes interact and the independence assumed by Eqs. (5)–(6) fails.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is a way of organizing an otherwise scattered literature: all LLM reasoning strategies live on a 2×2 grid whose axes are the fast/slow knowledge boundary (computation depth) and the internal/external knowledge boundary (knowledge source). Fast thinking is the direct, step-free answer; slow thinking is chain-of-thought-style deliberation and verification; tool-augmented thinking is any use of external tools, retrieval, or computation, and it can be paired with either fast or slow processing. The paper further claims that every existing adaptive-selection method—implicit ones learned through fine-tuning or reinforcement learning and explicit ones using rules or external routers—can be described by the same two-stage procedure: extract decision factors such as model confidence, task complexity, and utility gain, then cross a threshold on the relevant axis (Eqs. (5)–(6)). This unification lets the field compare methods by which factor they use and where the decision is made, rather than by surface differences in implementation.

Load-bearing premise

The grid assumes that how deeply an LLM reasons and whether it uses external knowledge are independent decisions; if tool use changes the optimal depth of reasoning, the two-by-two taxonomy misrepresents real systems.

Editorial extensions

If this is right

  • Tool-augmented thinking is demoted from a third mode to an axis: retrieval and tools can be combined with either fast direct answers or slow deliberative chains, making Fast-External—simple RAG and function calling—a recognized quadrant rather than an anomaly.
  • All selection-based adaptive-reasoning methods become comparable within one two-stage pipeline; a reader can ask separately what decision factors a method extracts and how it computes thresholds, which is a cleaner comparison than listing architectures.
  • Confidence, task complexity, and utility gain are complementary signals; the review implies that combining calibrated confidence with complexity estimates should give more reliable routing than any single factor.
  • The framework explicitly excludes fusion approaches such as SMaRT that ensemble multiple strategies; any method that aggregates rather than routes falls outside the proposed unification.

Reading between the lines

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

  • Editorial inference: if the two axes interact, the natural extension is to make the fast/slow threshold a function of the internal/external decision (feed $k_s$ into the computation of $\Phi_d$); the paper names this as future work, and a router trained on tool-availability as a depth feature would test it.
  • Editorial inference: the Fast-External quadrant may be more common in production than in the literature—simple RAG and API-calling systems already do exactly this—so the taxonomy predicts that research attention to this quadrant will grow as latency budgets tighten.
  • Editorial inference: the 2×2 grid suggests a benchmark design: vary required reasoning depth and required tool access independently across a task suite, then measure a selection policy's accuracy-latency frontier; policies that trace a convex frontier would confirm the two levers are worth tuning separately.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 7 minor

Summary. This review paper proposes a taxonomy of LLM reasoning strategies organized along two knowledge boundaries: a fast/slow boundary separating intuitive from deliberative computation, and an internal/external boundary separating parametric knowledge from tool-augmented knowledge. The authors formalize adaptive reasoning selection as a two-stage process of decision-factor extraction (Eqs. 1-2) and threshold-based strategy selection (Eqs. 5-6), and survey recent methods under this framework, grouping them into implicit and explicit selection paradigms and into decision factors of model confidence, task complexity, and utility gain. The paper concludes with future directions including pre-training integration, unified selection, orchestration, multimodal selection, personalization, and robustness. The central claim is that tool-augmented thinking is not a third mode but an external-knowledge dimension that intersects with the fast/slow dimension.

Significance. The taxonomy is conceptually appealing and timely, organizing a rapidly growing literature on adaptive reasoning in LLMs into an interpretable 2x2 structure. The claim that tool-augmented thinking should be treated as a knowledge-source dimension rather than a standalone mode is a useful corrective to the common three-way categorization. The survey covers a wide range of recent methods (roughly thirty) and includes worked examples for MUR, FLARE, and AdaptThink. The paper is honest about the exclusion of fusion-based methods and acknowledges potential interactions between the two dimensions. If the orthogonality concern is resolved, the framework could serve as a common vocabulary for future work. The main weaknesses are the unsubstantiated novelty claim and the unresolved status of dimension coupling in the formal model.

major comments (2)
  1. [Section 2, Eqs. (5)-(6)] The formalization treats reasoning-depth and knowledge-source decisions as independent threshold crossings with separate threshold vectors δd and δs. However, the paper itself concedes at the end of Section 2 that 'these two dimensions may still interact' and proposes only a future extension. This is load-bearing because the novelty claim rests on the orthogonality of the two boundaries. Interleaved methods surveyed in Section 5.2.2 (ReTool, R3-RAG, ARTIST) jointly optimize reasoning and retrieval, which cannot be faithfully represented by independent per-decision thresholds without an added coupling term. The authors should either provide evidence that such methods can be decomposed into independent threshold decisions per step, or extend Eqs. (5)-(6) with a coupling term (e.g., making δd a function of ks) and soften the 'orthogonal' wording. Without this, the descriptive adequacy of the taxonomy for tool-augmented slow thinking remains unresolved.
  2. [Introduction (contributions list)] The paper asserts that this is 'the first comprehensive review that systematically examines the adaptive selection among all three major reasoning strategies,' but it does not compare with prior surveys such as ref. [5] ('From System 1 to System 2: A Survey of Reasoning Large Language Models'). Because ref. [5] and other reasoning surveys exist, the novelty claim needs to be substantiated with an explicit discussion of what prior reviews cover and what this review adds. This is important because the contribution is framed around this novelty; if the claim is overstated, the contribution is weakened.
minor comments (7)
  1. [References [21] and [51]] The same FLARE paper appears as both [21] and [51]; these duplicate references should be consolidated.
  2. [Section 5.2.1 and Figure 3] The method for reference [28] is referred to as 'MeCo' in the text but 'MetaTrigger' in Figure 3; please unify the naming.
  3. [Figure 3] The spelling 'ARITIST' should be corrected to 'ARTIST'.
  4. [Section 5.1.1] The acronym 'CRL' for reference [62] is not defined and may be confusing; consider renaming or adding a definition.
  5. [Abstract and Section 2] The abstract uses 'two orthogonal knowledge boundaries,' but Section 2 later qualifies that the dimensions may interact; recommend replacing 'orthogonal' with 'distinct' or 'conceptually independent' to avoid overclaiming.
  6. [Table 1] Table 1 does not indicate which of the four quadrants (Fast-Internal, Slow-Internal, Fast-External, Slow-External) each method falls into; adding a quadrant column would make the taxonomy easier to use.
  7. [Section 2, Eq. (5)] There is a formatting issue in Eq. (5) where 'δ d,t' contains an extra space; please ensure the final typeset version is consistent.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the two-axis taxonomy is a definitional framework, not a derivation; self-citations are non-load-bearing examples.

full rationale

The paper is a survey/taxonomy and makes no empirical prediction. Section 2's Eqs. (5)-(6) are definitions of decision functions given thresholds; they do not derive any result from the taxonomy. The claimed orthogonality of the fast/slow and internal/external axes is an organizing assumption, not a fitted or self-cited premise. The paper explicitly acknowledges at the end of Section 2 that 'these two dimensions may still interact' and that explicit coupling is 'underexplored,' and it offers a future extension; this is a stated limitation rather than a hidden circularity. The three self-references ([28], [52], [68]) are used only as surveyed examples or as an application citation; none is load-bearing for the taxonomy or for any derivation. No parameter is fitted and no prediction is made from a fitted input. The closest concern—that the 2x2 grid is definitional—does not meet the standard of circularity because the paper does not claim to derive the grid from evidence; it proposes it as an organizing framework. Therefore, no circular step can be quoted with a specific reduction, and the honest finding is no significant circularity.

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

No numerical parameters are fitted in this review; thresholds and decision scores in Eqs. (5)-(6) belong to the surveyed methods rather than to the paper's own claims. The 'knowledge boundaries' are descriptive categories, not invented entities with independent falsifiable predictions.

assumptions (4)
  • domain assumption The fast/slow and internal/external dimensions are orthogonal, forming a 2x2 grid of reasoning strategies.
    Central premise of Figure 1 and Section 5; the paper undercuts it in Section 2 by noting that tool use and reasoning depth interact.
  • ad hoc to paper Any selection-based adaptive reasoning method can be decomposed into decision factor extraction (Eqs. 1-2) and score-threshold selection (Eqs. 3-6).
    Section 2 imposes this abstraction; the paper maps RL policies, routers, and confidence heuristics onto it, but the mapping is interpretive and some methods, like AdaptThink, have no explicit threshold.
  • domain assumption The works cited in Figure 3 and Table 1 are representative of the field and sufficient to support the taxonomy.
    No systematic search or inclusion criteria are given, so coverage is assumed.
  • domain assumption Cognitive psychology concepts such as System 1/System 2 and metacognition transfer to LLMs as a useful analogy.
    Section 3 explicitly says the analogy is pragmatic rather than a claim of cognitive equivalence, which limits but does not remove reliance on the analogy for framing.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Fast, Slow, and Tool-augmented Thinking for LLMs: A Review." pith.science (2026). https://pith.science/paper/OPLFKFBY

@misc{pith2026250812265,
  author       = {Pith},
  title        = {Pith review of: Fast, Slow, and Tool-augmented Thinking for LLMs: A Review},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/OPLFKFBY}},
  note         = {Machine review of arXiv:2508.12265}
}
read the original abstract

Large Language Models (LLMs) have demonstrated remarkable progress in reasoning across diverse domains. However, effective reasoning in real-world tasks requires adapting the reasoning strategy to the demands of the problem, ranging from fast, intuitive responses to deliberate, step-by-step reasoning and tool-augmented thinking. Drawing inspiration from cognitive psychology, we propose a novel taxonomy of LLM reasoning strategies along two knowledge boundaries: a fast/slow boundary separating intuitive from deliberative processes, and an internal/external boundary distinguishing reasoning grounded in the model's parameters from reasoning augmented by external tools. We systematically survey recent work on adaptive reasoning in LLMs and categorize methods based on key decision factors. We conclude by highlighting open challenges and future directions toward more adaptive, efficient, and reliable LLMs.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

71 extracted references · 27 canonical work pages

  1. [5]

    From system 1 to system 2: A survey of reasoning large language models, 2025

    Li Z Z, Zhang D, Zhang M L, Zhang J, Liu Z, Yao Y , Xu H, Zheng J, Wang P J, Chen X, others . From system 1 to system 2: A survey of reasoning large language models, 2025. URL https://arxiv. org/abs/2502.17419, 2025

  2. [1]

    Claude sonnet 4, 2025

    Anthropic . Claude sonnet 4, 2025

  3. [2]

    Deepseek-r1: Incen- tivizing reasoning capability in llms via reinforcement learning

    Guo D, Yang D, Zhang H, Song J, Zhang R, Xu R, Zhu Q, Ma S, Wang P, Bi X, others . Deepseek-r1: Incen- tivizing reasoning capability in llms via reinforcement learning. arXiv preprint arXiv:2501.12948, 2025

  4. [3]

    Openai o3, 2025

    OpenAI . Openai o3, 2025

  5. [4]

    Grok 4, 2025 18 Front

    xAI . Grok 4, 2025 18 Front. Comput. Sci., 2026, 0(0): 1–20

  6. [6]

    Thinking, fast and slow

    Kahneman D. Thinking, fast and slow. macmillan, 2011

  7. [7]

    Dy- nathink: Fast or slow? a dynamic decision-making framework for large language models

    Pan J, Zhang Y , Zhang C, Liu Z, Wang H, Li H. Dy- nathink: Fast or slow? a dynamic decision-making framework for large language models. arXiv preprint arXiv:2407.01009, 2024

  8. [8]

    Chain-of-thought prompting elicits reasoning in large language models

    Wei J, Wang X, Schuurmans D, Bosma M, Xia F, Chi E, Le Q V , Zhou D, others . Chain-of-thought prompting elicits reasoning in large language models. Advances in neural information processing systems, 2022, 35: 24824– 24837

Show all 71 references
  1. [9]

    Tree of thoughts: Deliberate problem solving with large language models

    Yao S, Yu D, Zhao J, Shafran I, Griffiths T, Cao Y , Narasimhan K. Tree of thoughts: Deliberate problem solving with large language models. Advances in neu- ral information processing systems, 2023, 36: 11809– 11822

  2. [10]

    Towards mitigat- ing llm hallucination via self reflection

    Ji Z, Yu T, Xu Y , Lee N, Ishii E, Fung P. Towards mitigat- ing llm hallucination via self reflection. In: Findings of the Association for Computational Linguistics: EMNLP

  3. [11]

    Im- proving llm reasoning through scaling inference com- putation with collaborative verification

    Liang Z, Liu Y , Niu T, Zhang X, Zhou Y , Yavuz S. Im- proving llm reasoning through scaling inference com- putation with collaborative verification. arXiv preprint arXiv:2410.05318, 2024

  4. [12]

    Let llms break free from overthinking via self-braking tuning

    Zhao H, Yan Y , Shen Y , Xu H, Zhang W, Song K, Shao J, Lu W, Xiao J, Zhuang Y . Let llms break free from overthinking via self-braking tuning. arXiv preprint arXiv:2505.14604, 2025

  5. [13]

    Large language models are zero-shot reasoners

    Kojima T, Gu S S, Reid M, Matsuo Y , Iwasawa Y . Large language models are zero-shot reasoners. Advances in neural information processing systems, 2022, 35: 22199– 22213

  6. [14]

    Detec- tion and mitigation of hallucination in large reason- ing models: A mechanistic perspective

    Sun Z, Wang Q, Wang H, Zhang X, Xu J. Detec- tion and mitigation of hallucination in large reason- ing models: A mechanistic perspective. arXiv preprint arXiv:2505.12886, 2025

  7. [15]

    Confident adaptive language modeling, 2022

    Schuster T, Fisch A, Gupta J, Dehghani M, Bahri D, Tran V Q, Tay Y , Metzler D. Confident adaptive language modeling, 2022

  8. [16]

    How much knowledge can you pack into the parameters of a language model? arXiv preprint arXiv:2002.08910, 2020

    Roberts A, Raffel C, Shazeer N. How much knowledge can you pack into the parameters of a language model? arXiv preprint arXiv:2002.08910, 2020

  9. [17]

    Adaptive retrieval without self-knowledge? bringing uncertainty back home

    Moskvoretskii V , Lysyuk M, Salnikov M, Ivanov N, Pletenev S, Galimzianova D, Krayko N, Konovalov V , Nikishina I, Panchenko A. Adaptive retrieval without self-knowledge? bringing uncertainty back home. arXiv preprint arXiv:2501.12835, 2025

  10. [18]

    Self-route: Automatic mode switching via capa- bility estimation for efficient reasoning

    He Y , Ding X, Cai B, Zhang Y , Xiong K, Sun Z, Qin B, Liu T. Self-route: Automatic mode switching via capa- bility estimation for efficient reasoning. arXiv preprint arXiv:2505.20664, 2025

  11. [19]

    Adacot: Pareto-optimal adaptive chain- of-thought triggering via reinforcement learning

    Lou C, Sun Z, Liang X, Qu M, Shen W, Wang W, Li Y , Yang Q, Wu S. Adacot: Pareto-optimal adaptive chain- of-thought triggering via reinforcement learning. arXiv preprint arXiv:2505.11896, 2025

  12. [20]

    Thinkswitcher: When to think hard, when to think fast

    Liang G, Zhong L, Yang Z, Quan X. Thinkswitcher: When to think hard, when to think fast. arXiv preprint arXiv:2505.14183, 2025

  13. [21]

    Active retrieval augmented generation

    Jiang Z, Xu F F, Gao L, Sun Z, Liu Q, Dwivedi-Yu J, Yang Y , Callan J, Neubig G. Active retrieval augmented generation. In: Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing. 2023, 7969–7992

  14. [22]

    Mur: Momentum uncer- tainty guided reasoning for large language models

    Yan H, Xu F, Xu R, Li Y , Zhang J, Luo H, Wu X, Tuan L A, Zhao H, Lin Q, others . Mur: Momentum uncer- tainty guided reasoning for large language models. arXiv preprint arXiv:2507.14958, 2025

  15. [23]

    Adaptthink: Rea- soning models can learn when to think

    Zhang J, Lin N, Hou L, Feng L, Li J. Adaptthink: Rea- soning models can learn when to think. arXiv preprint arXiv:2505.13417, 2025

  16. [24]

    Learning when to think: Shaping adaptive reasoning in r1-style models via multi-stage rl

    Tu S, Lin J, Zhang Q, Tian X, Li L, Lan X, Zhao D. Learning when to think: Shaping adaptive reasoning in r1-style models via multi-stage rl. arXiv preprint arXiv:2505.10832, 2025

  17. [25]

    Certainty- guided reasoning in large language models: A dy- namic thinking budget approach

    Nogueira J P, Sun W, Silva A, Zumot L. Certainty- guided reasoning in large language models: A dy- namic thinking budget approach. arXiv preprint arXiv:2509.07820, 2025

  18. [26]

    Dynamic early exit in reasoning models

    Yang C, Si Q, Duan Y , Zhu Z, Zhu C, Li Q, Lin Z, Cao L, Wang W. Dynamic early exit in reasoning models. arXiv preprint arXiv:2504.15895, 2025

  19. [27]

    Seakr: Self-aware knowledge retrieval for adaptive re- trieval augmented generation, 2024

    Yao Z, Qi W, Pan L, Cao S, Hu L, Liu W, Hou L, Li J. Seakr: Self-aware knowledge retrieval for adaptive re- trieval augmented generation, 2024

  20. [28]

    Adaptive tool use in large language models with meta-cognition trigger

    Li W, Li D, Dong K, Zhang C, Zhang H, Liu W, Wang Y , Tang R, Liu Y . Adaptive tool use in large language models with meta-cognition trigger. arXiv preprint arXiv:2502.12961, 2025

  21. [29]

    Adaptive retrieval- augmented generation for conversational systems

    Wang X, Sen P, Li R, Yilmaz E. Adaptive retrieval- augmented generation for conversational systems. arXiv preprint arXiv:2407.21712, 2024

  22. [30]

    Uncertainty-guided chain-of-thought for code genera- tion with llms

    Zhu Y , Li G, Jiang X, Li J, Mei H, Jin Z, Dong Y . Uncertainty-guided chain-of-thought for code genera- tion with llms. arXiv preprint arXiv:2503.15341, 2025

  23. [31]

    Reasoning models know when they’re right: Prob- ing hidden states for self-verification

    Zhang A, Chen Y , Pan J, Zhao C, Panda A, Li J, He H. Reasoning models know when they’re right: Prob- ing hidden states for self-verification. arXiv preprint arXiv:2504.05419, 2025 Xinda Jia et al. Fast, Slow, and Tool-augmented Thinking for LLMs: A Review 19

  24. [32]

    Sugar: Leveraging contextual confidence for smarter retrieval

    Zubkova H, Park J H, Lee S W. Sugar: Leveraging contextual confidence for smarter retrieval. In: ICASSP 2025-2025 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). 2025, 1–5

  25. [33]

    Smart: Select, mix, and reinvent–a strategy fusion framework for llm-driven reasoning and planning

    Verma N, Bharadwaj M, Jang W, Singh H, Wang Y , Fashandi H, Lee C. Smart: Select, mix, and reinvent–a strategy fusion framework for llm-driven reasoning and planning. arXiv preprint arXiv:2510.18095, 2025

  26. [34]

    Metacognition and cognitive monitoring: A new area of cognitive–developmental inquiry

    Flavell J H. Metacognition and cognitive monitoring: A new area of cognitive–developmental inquiry. American psychologist, 1979, 34(10): 906

  27. [35]

    Cognitive load during problem solving: Ef- fects on learning

    Sweller J. Cognitive load during problem solving: Ef- fects on learning. Cognitive science, 1988, 12(2): 257– 285

  28. [36]

    Au- tol2s: Auto long-short reasoning for efficient large lan- guage models

    Luo F, Chuang Y N, Wang G, Le H A D, Zhong S, Liu H, Yuan J, Sui Y , Braverman V , Chaudhary V , others . Au- tol2s: Auto long-short reasoning for efficient large lan- guage models. arXiv preprint arXiv:2505.22662, 2025

  29. [37]

    Tl; dr: Too long, do re-weighting for effcient llm reasoning compression

    Li Z Z, Liang X, Tang Z, Ji L, Wang P, Xu H, Huang H, Deng W, Wu Y N, Gong Y , others . Tl; dr: Too long, do re-weighting for effcient llm reasoning compression. arXiv preprint arXiv:2506.02678, 2025

  30. [38]

    Self-rag: Learning to retrieve, generate, and critique through self- reflection

    Asai A, Wu Z, Wang Y , Sil A, Hajishirzi H. Self-rag: Learning to retrieve, generate, and critique through self- reflection. In: The Twelfth International Conference on Learning Representations. 2023

  31. [39]

    Toolformer: Language models can teach themselves to use tools,

    Schick T, Dwivedi-Yu J, Dessì R, Raileanu R, Lomeli M, Zettlemoyer L, Cancedda N, Scialom T. Toolformer: Language models can teach themselves to use tools,

  32. [40]

    We- bgpt: Browser-assisted question-answering with human feedback

    Nakano R, Hilton J, Balaji S, Wu J, Ouyang L, Kim C, Hesse C, Jain S, Kosaraju V , Saunders W, others . We- bgpt: Browser-assisted question-answering with human feedback. arXiv preprint arXiv:2112.09332, 2021

  33. [41]

    Synapseroute: An auto- route switching framework on dual-state large language model

    Zhang W, Qiao S, Luo L, Li Y , Zheng C, Xu Q, Li M, Gui Y , He Y , Qiu J, others . Synapseroute: An auto- route switching framework on dual-state large language model. arXiv preprint arXiv:2507.02822, 2025

  34. [42]

    Kat-v1: Kwai-autothink technical report

    Zhan Z, Deng K, Tang H, Xiang W, Wu K, Li W, Zhu W, Xu J, Huang L, Feng Z, others . Kat-v1: Kwai-autothink technical report. arXiv preprint arXiv:2507.08297, 2025

  35. [43]

    Think only when you need with large hybrid-reasoning models

    Jiang L, Wu X, Huang S, Dong Q, Chi Z, Dong L, Zhang X, Lv T, Cui L, Wei F. Think only when you need with large hybrid-reasoning models. arXiv preprint arXiv:2505.14631, 2025

  36. [44]

    To trust or not to trust? enhancing large language models’ situ- ated faithfulness to external contexts

    Huang Y , Chen S, Cai H, Dhingra B. To trust or not to trust? enhancing large language models’ situ- ated faithfulness to external contexts. arXiv preprint arXiv:2410.14675, 2024

  37. [45]

    Exaone 4.0: Unified large language models integrating non-reasoning and reasoning modes

    Research L, Bae K, Choi E, Choi K, Choi S J, Choi Y , Han K, Hong S, Hwang J, Hwang T, others . Exaone 4.0: Unified large language models integrating non-reasoning and reasoning modes. arXiv preprint arXiv:2507.11407, 2025

  38. [46]

    Introducing gpt-5, 2025

    OpenAI . Introducing gpt-5, 2025

  39. [47]

    Thinking with nothinking calibration: A new in-context learn- ing paradigm in reasoning large language models

    Wu H, Xu B, Shu Y , Yang M, Qin C. Thinking with nothinking calibration: A new in-context learn- ing paradigm in reasoning large language models. arXiv preprint arXiv:2508.03363, 2025

  40. [48]

    Z1: Efficient test-time scaling with code

    Yu Z, Wu Y , Zhao Y , Cohan A, Zhang X P. Z1: Efficient test-time scaling with code. arXiv preprint arXiv:2504.00810, 2025

  41. [49]

    Qwen3 technical report

    Yang A, Li A, Yang B, Zhang B, Hui B, Zheng B, Yu B, Gao C, Huang C, Lv C, others . Qwen3 technical report. arXiv preprint arXiv:2505.09388, 2025

  42. [50]

    Llama-nemotron: Efficient reasoning models

    Bercovich A, Levy I, Golan I, Dabbah M, El-Yaniv R, Puny O, Galil I, Moshe Z, Ronen T, Nabwani N, others . Llama-nemotron: Efficient reasoning models. arXiv preprint arXiv:2505.00949, 2025

  43. [51]

    Active retrieval augmented gener- ation

    Jiang Z, Xu F, Gao L, Sun Z, Liu Q, Dwivedi-Yu J, Yang Y , Callan J, Neubig G. Active retrieval augmented gener- ation. In: Bouamor H, Pino J, Bali K, eds, Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing. December 2023

  44. [52]

    The web can be your oyster for improving large language models

    Li J, Tang T, Zhao W X, Wang J, Nie J Y , Wen J R. The web can be your oyster for improving large language models. arXiv preprint arXiv:2305.10998, 2023

  45. [53]

    Dragin: Dynamic retrieval augmented generation based on the information needs of large language models, 2024

    Su W, Tang Y , Ai Q, Wu Z, Liu Y . Dragin: Dynamic retrieval augmented generation based on the information needs of large language models, 2024

  46. [54]

    Retrieve only when it needs: Adaptive retrieval augmentation for hallucination mitigation in large language models

    Ding H, Pang L, Wei Z, Shen H, Cheng X. Retrieve only when it needs: Adaptive retrieval augmentation for hallucination mitigation in large language models. arXiv preprint arXiv:2402.10612, 2024

  47. [55]

    Alignment for efficient tool calling of large language models

    Xu H, Wang Z, Zhu Z, Pan L, Chen X, Chen L, Yu K. Alignment for efficient tool calling of large language models. arXiv preprint arXiv:2503.06708, 2025

  48. [56]

    Adapting while learning: Ground- ing llms for scientific problems with intelligent tool us- age adaptation

    Lyu B, Cao Y , Watson-Parris D, Bergen L, Berg- Kirkpatrick T, Yu R. Adapting while learning: Ground- ing llms for scientific problems with intelligent tool us- age adaptation. arXiv preprint arXiv:2411.00412, 2024

  49. [57]

    To code or not to code? adaptive tool integration for math language models via expectation-maximization

    Wang H, Li L, Qu C, Zhu F, Xu W, Chu W, Lin F. To code or not to code? adaptive tool integration for math language models via expectation-maximization. arXiv preprint arXiv:2502.00691, 2025

  50. [58]

    Retool: Reinforcement learning for strategic tool use in llms

    Feng J, Huang S, Qu X, Zhang G, Qin Y , Zhong B, Jiang C, Chi J, Zhong W. Retool: Reinforcement learning for strategic tool use in llms. arXiv preprint arXiv:2504.11536, 2025

  51. [59]

    R3-rag: Learning step-by- step reasoning and retrieval for llms via reinforcement 20 Front

    Li Y , Luo Q, Li X, Li B, Cheng Q, Wang B, Zheng Y , Wang Y , Yin Z, Qiu X. R3-rag: Learning step-by- step reasoning and retrieval for llms via reinforcement 20 Front. Comput. Sci., 2026, 0(0): 1–20 learning. arXiv preprint arXiv:2505.23794, 2025

  52. [60]

    Agentic reasoning and tool integration for llms via reinforcement learning

    Singh J, Magazine R, Pandya Y , Nambi A. Agentic reasoning and tool integration for llms via reinforcement learning. arXiv preprint arXiv:2505.01441, 2025

  53. [61]

    Toolken+: Improving llm tool usage with reranking and a reject option

    Yakovlev K, Nikolenko S, Bout A. Toolken+: Improving llm tool usage with reranking and a reject option. arXiv preprint arXiv:2410.12004, 2024

  54. [62]

    Confidence in the reasoning of large language models

    Pawitan Y , Holmes C. Confidence in the reasoning of large language models. Harvard Data Science Review, 2025, 7(1)

  55. [63]

    Self-consistency improves chain of thought reasoning in language models

    Wang X, Wei J, Schuurmans D, Le Q, Chi E, Narang S, Chowdhery A, Zhou D. Self-consistency improves chain of thought reasoning in language models. arXiv preprint arXiv:2203.11171, 2022

  56. [64]

    Wizardlm: Empowering large pre- trained language models to follow complex instructions

    Xu C, Sun Q, Zheng K, Geng X, Zhao P, Feng J, Tao C, Lin Q, Jiang D. Wizardlm: Empowering large pre- trained language models to follow complex instructions. In: The Twelfth International Conference on Learning Representations. 2024

  57. [65]

    Gaia: a benchmark for general ai assistants, 2023

    Mialon G, Fourrier C, Swift C, Wolf T, LeCun Y , Scialom T. Gaia: a benchmark for general ai assistants, 2023

  58. [66]

    When do llms need retrieval augmentation? mitigating llms’ overconfidence helps retrieval augmentation

    Ni S, Bi K, Guo J, Cheng X. When do llms need retrieval augmentation? mitigating llms’ overconfidence helps retrieval augmentation. In: Findings of the Association for Computational Linguistics: ACL 2024. 2024, 11375– 11388

  59. [67]

    A survey on hallucination in large language models: Principles, taxonomy, challenges, and open questions

    Huang L, Yu W, Ma W, Zhong W, Feng Z, Wang H, Chen Q, Peng W, Feng X, Qin B, others . A survey on hallucination in large language models: Principles, taxonomy, challenges, and open questions. ACM Trans- actions on Information Systems, 2025, 43(2): 1–55

  60. [68]

    The real barrier to llm agent usability is agentic roi

    Liu W, Qin J, Huang X, Zeng X, Xi Y , Lin J, Wu C, Wang Y , Shang L, Tang R, others . The real barrier to llm agent usability is agentic roi. arXiv preprint arXiv:2505.17767, 2025

  61. [69]

    Multi-agent collaboration mechanisms: A survey of llms

    Tran K T, Dao D, Nguyen M D, Pham Q V , O’Sullivan B, Nguyen H D. Multi-agent collaboration mechanisms: A survey of llms. arXiv preprint arXiv:2501.06322, 2025

  62. [70]

    The interaction between text modality and the learner’s modality preference influ- ences comprehension and cognitive load

    Lehmann J, Seufert T. The interaction between text modality and the learner’s modality preference influ- ences comprehension and cognitive load. Frontiers in Psychology, 2020, 10: 2820

  63. [2023]

    arXiv preprint arXiv:2302.04761, 2023

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.