Pith. sign in

REVIEW 5 major objections 5 minor 3 cited by

Less is More: Optimizing Function Calling for LLM Execution on Edge Devices

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 Providing an LLM with fewer, better-chosen API tools improves function-calling accuracy while cutting execution time and power on edge devices.

desk verdict A practical tool-selection scheme with plausible results, but the evaluation conflates tool pruning with context shortening and doesn't isolate why it helps. read the letter →

arxiv 2411.15399 v1 pith:OOVSBFOW submitted 2024-11-23 cs.PF cs.DCcs.LG

Varatheepan Paramanayakam , Andreas Karatzas , Iraklis Anagnostopoulos , Dimitrios Stamoulis This is my paper · ORCID
classification cs.PFcs.DCcs.LG
keywords functioncallingedgeAIdynamictoolselectionLLMagentslatent-spaceretrievalquantizedLLMsFAISSrecommendation
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

The paper argues that handing an LLM agent every available API at once is self-defeating on edge devices: with 46 tools in context, a quantized 8B model fails a simple mapping query, and succeeds once the tool list is cut to 19. It introduces Less-is-More, a fine-tuning-free pipeline that at runtime asks the LLM to describe the tools it would need, embeds those descriptions, retrieves the closest real tools from offline-built search levels, and only then calls the function with the reduced tool set. The reported effect is that success rate and tool accuracy rise across several quantized LLMs while execution time drops by up to 70-80% and power consumption by up to 40-45%. If the effect holds, small on-device models can act as capable agents without retraining.

What carries the argument

The load-bearing object is a three-level latent tool space built offline with the MPNet sentence encoder: Search Level 1 embeds each API description individually, Search Level 2 augments the space with GPT-generated variations of benchmark queries and agglomerative clusters the results so that tools used in the same task lie close together, and Search Level 3 is simply the full JSON tool set used as fallback. At runtime the same encoder embeds the LLM's self-generated ideal tool descriptions plus the user query, and a FAISS k-NN search retrieves the top-k items from Levels 1 and 2; the Controller picks the level with the higher average similarity and passes only that reduced tool set to the agent. This lets the method trade tool granularity against query complexity without fine-tuning.

What would settle it

Run the same edge models on the same benchmarks with the same reduced context window but replace the LLM-generated ideal descriptions with the user query's own embedding as the retrieval key; if success rate stays the same, the Recommender step adds nothing and the gains come purely from fewer tools. Alternatively, measure success rate when the reduced tool set is chosen at random instead of by similarity: random selection performing equally would show that tool-space size, not relevance, drives the result.

Watch

Extended reading notes

Core claim

Less-is-More claims that selectively reducing the number of tools available to an LLM is itself a performance optimization: fewer options reduce the model's confusion, let it select and invoke the right function more often, and shrink the context window enough to cut latency and energy. The discovery is a mechanism, not just a heuristic: a runtime Recommender generates ideal tool descriptions from the user query alone, and a Controller matches them through k-NN similarity against three offline-constructed representations of the real tool space, namely individual tools, clusters of tools that tend to be used together, and the full set as fallback. Across six LLMs and two benchmarks, the reduced-tool protocol improves success rate and tool accuracy over both the default all-tools setup and a similarity-selection baseline, with the largest gains on the sequential GeoEngine tasks.

Load-bearing premise

The load-bearing premise is that an LLM prompted with no API list can still describe the tools a task needs accurately enough that embedding-based similarity retrieves the real tools; if the ideal descriptions are off, the reduced set is wrong and the claimed accuracy gains vanish.

Editorial extensions

If this is right

  • Off-the-shelf quantized LLMs can serve as function-calling agents on edge boards without fine-tuning, widening the range of deployable models.
  • Smaller context windows become sufficient, which compounds the latency and power savings beyond the tool-count reduction alone.
  • The scheme is plug-and-play: any existing LLM with function calling can be wrapped by the Recommender and Controller without retraining or weight changes.
  • Applications that must keep user data on-device for privacy gain a practical path to cloud-free agent execution.
  • Even when accuracy gains are small, as with Mistral on BFCL, the method still delivers large execution-time and power reductions.

Reading between the lines

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

  • The less-is-more effect likely generalizes beyond tool calling: any LLM task with a large enumerated option set, such as multiple-choice, routing, or retrieval, may suffer similar confusion, and the same reduction-by-retrieval pattern could be tested there.
  • A cheaper variant can be tested that skips the LLM Recommender and embeds the raw query directly; if accuracy holds, the entire Recommender call is removable, saving its latency and power.
  • Search Level 2 clusters are built from GPT-generated query variations, so generalization to new tool sets could depend on how well those synthetic queries cover real usage patterns; usage logs from deployed agents would be a stronger clustering signal.
  • The 0.5 similarity threshold for falling back to all tools is a tunable knob; lowering it trades more Recommender mistakes for a smaller context, and mapping that tradeoff could produce an adaptive per-model policy.
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 proposes Less-is-More, a fine-tuning-free dynamic tool-selection scheme for LLM function calling on edge devices. API descriptions and GPT-4-augmented benchmark queries are embedded offline into latent spaces at three granularities (individual tools, tool clusters, full tool set). At runtime, an LLM is first asked to describe the 'ideal' tools for the user query without seeing any API definitions; a FAISS k-NN search then retrieves the most similar real tools or clusters, and the agent is invoked with only that reduced tool set. Experiments on BFCL and GeoEngine with six quantized LLMs on an NVIDIA Jetson AGX Orin report improved success rates and tool accuracy, with execution time reduced by up to 70-80% and power by up to 40%.

Significance. The problem is timely: deploying capable function-calling agents on resource-constrained edge hardware is an active research area, and a plug-and-play, fine-tuning-free tool-selection method would be practically valuable. The paper has concrete strengths: direct measurements on edge hardware, multiple LLMs and quantization variants, comparison against a prior similarity-based baseline (Gorilla), and a hierarchical search-level design that is simple to implement. However, the central empirical claim is not currently established because the experiments confound tool selection with context-window shortening, the universal-improvement conclusion is contradicted by the paper's own reported Mistral results, and part of the evaluation is built from benchmark-derived clusters and selective baselines. If these issues are addressed, the contribution would be solid; in its present form, the significance is uncertain.

major comments (5)
  1. [§IV, paragraph beginning 'We determined the minimum context window...'] The comparison varies three factors at once. The default baseline uses the full tool set at a 16k context window, while Less-is-More and Gorilla use k=3 or k=5 tools at an 8k context window. Consequently, the reported execution-time reductions (up to 70-80%) and power reductions cannot be attributed to tool selection: prefill cost scales with prompt length, so the 16k-to-8k context cut is a plausible dominant factor. The success-rate claim is likewise underdetermined. Without a same-context non-selective subset baseline (e.g., random or fixed k-tool subset at 8k context) or a full-tool-set condition at 8k context, the paper cannot distinguish 'fewer tools help', 'more relevant tools help', and 'shorter prompts help'. The Table II 16K 46-tool vs 16K 19-tool illustration is a single query, not a statistical comparison. I request at least two additional conditions: (a) all tools at 8k context and (b) an arbitrary k-tool subset at 8k context, with absolute execution time and power reported.
  2. [§IV, Mistral-8b BFCL paragraph and the 'Overall' paragraph] The text states for Mistral-8b on BFCL that 'the optimizations did not result in any gain in success rate and tool accuracy', yet the next paragraphs conclude that 'our method improved all four metrics—success rate, tool accuracy, execution time, and power consumption—across all the LLMs'. On GeoEngine, Mistral-8b is reported as having 'Execution time was 10% higher for some variations', which further contradicts a universal-improvement claim. The authors should restrict the conclusions to the metrics and models for which gains are actually observed, or present the negative cases with equal prominence.
  3. [§IV, GeoEngine paragraph] Phi3 and Qwen2-1.5b are excluded from the GeoEngine analysis because their default success rate was approximately 10% and their execution-time/power measurements were deemed unreliable. Excluding the two models with the weakest default performance from the more complex benchmark removes the cases where the method is most likely to fail and makes the cross-model claim unrepresentative. Please include these models in a secondary analysis, or explicitly limit the GeoEngine claim to the reported subset of models.
  4. [§III-B] The Tool Recommender is prompted to produce 'ideal' tool descriptions without being shown any API definitions. The entire retrieval chain depends on this step: if the base quantized LLM cannot infer plausible tool descriptions for a domain such as GeoEngine's remote-sensing tools, the similarity search cannot retrieve relevant tools. This premise is not validated. The paper reports no ablation comparing the LLM-generated descriptions against ground-truth tool descriptions, no comparison against directly embedding the user query, and no oracle-retrieval upper bound. In addition, the free parameters k and the 0.5 fallback threshold are not subjected to sensitivity analysis. These experiments are needed before the success-rate gains can be attributed to the recommender.
  5. [§III-A, Search Level 2] The augmented latent space is constructed from GPT-4-augmented queries sampled from the same GeoEngine and BFCL categories used later for evaluation. Although the paper says these generated queries 'are not used for training or runtime decisions', they are used to form the clusters that the Tool Controller selects between at runtime, so the disclaimer is misleading. This is an evaluation-contamination risk: the cluster structure may encode benchmark-specific query-tool associations, and the reported success rates may not transfer to unseen tool spaces. I request a leave-one-benchmark-out or held-out-category experiment, or a demonstration that clusters built without the benchmark queries yield similar performance.
minor comments (5)
  1. [§III-A and §III-B] The latent space is described as 768-dimensional in one place and '728-dimensional' in another; please reconcile the inconsistency.
  2. [§IV, Figures 2 and 3] Only normalized execution time and normalized power are reported; without absolute values, units, or confidence intervals/k-fold repetitions, the magnitude of the efficiency gains is hard to assess.
  3. [§IV] The sentence 'For the default models, we also tested context windows larger than 16k' is not backed by any presented result; either include the results or remove the claim.
  4. [Table II] The last row 'Max drop ↓ 43% ↓ 19%' is ambiguous: it is unclear what quantity is maximized and why the row is labeled a drop; please define the comparison explicitly.
  5. [§III-C] The statement that for simple queries the LLM 'would likely recommend just one ideal tool description' is informal; the paper does not analyze the distribution of the number of recommended tools, which is relevant for choosing k.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: LiS is an empirical pipeline evaluated against external benchmarks; self-citations and benchmark-derived clustering do not make the reported gains equivalent to the method's inputs.

full rationale

The paper makes no formal derivation that could reduce to its inputs; the central claims are experimental. The Tool Recommender/Controller pipeline (Sec. III-B/C) is an ordinary LLM-prompting-plus-embedding-similarity scheme, and its success is not guaranteed by construction: the LLM could generate poor ideal tool descriptions, and the k-NN search could retrieve irrelevant tools, which would lower success rates. The only step that approaches circularity is the offline clustering in Sec. III-A, where GPT-4-augmented queries are sampled from the training categories of the very benchmarks (BFCL, GeoEngine) used for evaluation; this risks mild evaluation contamination, but the text states the augmented queries are used only as 'noisy' clustering inputs and are not used for training or runtime decisions, and the final success-rate measurements are not computed from those cluster assignments. The paper's self-citations (GeoEngine [12], LLM-DCache [27], LLM-tool compiler [32], GeckOpt [34], GeoLLM evaluation [41], Pythia [2]) are used as related work, benchmark, or prior-context references and are not invoked to forbid alternatives or to justify the key insight. Similarly, no parameter is fitted to the reported success rates and then renamed a prediction: k is chosen (3 or 5) and the Search Level is selected by similarity scores, not by optimizing the benchmark metrics. Confounds such as context-window reduction (16K vs 8K) and the absence of a non-selective k-tool baseline are real experimental-design threats, but they are correctness/attribution issues, not circularity: the reported execution-time and power reductions do not reduce by definition to the method's parameters. Overall, no circular step can be exhibited from the text.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The method depends on the quality of MPNet embeddings, on the LLM's ability to invent ideal tool descriptions without seeing real tools, on GPT-4-augmented benchmark queries producing clusters that generalize to test queries, and on k and the fallback threshold. No invented physical entities are introduced.

free parameters (5)
  • k (number of retrieved tools or clusters) = k=3,5
    Controller runs k-NN and top-k selection; no selection rule for k and no sensitivity analysis beyond two values.
  • Fallback threshold = 0.5
    If average top-k similarity scores are below 0.5 the controller falls back to all tools; the threshold is presented without justification or sensitivity analysis.
  • GPT-4 generated queries per category = 10
    Clustering augmentation samples 10 queries per benchmark category; the number appears arbitrary.
  • ROUGE similarity threshold for augmented queries = Not stated
    Generated queries are filtered by ROUGE score to ensure diversity, but the actual cutoff is not reported.
  • Number of agglomerative clusters = Not stated
    The clustering cutoff for Search Level 2 is not reported; it determines the coarseness of tool groups and is chosen offline.
assumptions (4)
  • domain assumption Semantic similarity in the MPNet embedding space is a reliable proxy for tool relevance.
    Section III-A uses embedding closeness to match LLM-generated ideal tool descriptions to real tools; no validation of the embedding's reliability for function-calling relevance.
  • domain assumption The LLM can generate accurate ideal tool descriptions when prompted with no actual tools.
    Section III-B feeds the query to the LLM with no APIs and relies on its self-generated tool descriptions; if the LLM cannot do this the retrieval fails.
  • domain assumption GPT-4-augmented queries from benchmark categories capture real tool-usage co-occurrence patterns.
    Section III-A builds clusters from these generated queries; the approach assumes the generated task permutations reflect real sequential tool combinations.
  • domain assumption Reducing the number of tools reduces LLM confusion and improves decision accuracy.
    This is the paper's key premise, supported only by the anecdote in Table II and aggregate benchmark results, not by a controlled ablation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Less is More: Optimizing Function Calling for LLM Execution on Edge Devices." pith.science (2026). https://pith.science/paper/OOVSBFOW

@misc{pith2026241115399,
  author       = {Pith},
  title        = {Pith review of: Less is More: Optimizing Function Calling for LLM Execution on Edge Devices},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/OOVSBFOW}},
  note         = {Machine review of arXiv:2411.15399}
}
read the original abstract

The advanced function-calling capabilities of foundation models open up new possibilities for deploying agents to perform complex API tasks. However, managing large amounts of data and interacting with numerous APIs makes function calling hardware-intensive and costly, especially on edge devices. Current Large Language Models (LLMs) struggle with function calling at the edge because they cannot handle complex inputs or manage multiple tools effectively. This results in low task-completion accuracy, increased delays, and higher power consumption. In this work, we introduce Less-is-More, a novel fine-tuning-free function-calling scheme for dynamic tool selection. Our approach is based on the key insight that selectively reducing the number of tools available to LLMs significantly improves their function-calling performance, execution time, and power efficiency on edge devices. Experimental results with state-of-the-art LLMs on edge hardware show agentic success rate improvements, with execution time reduced by up to 70% and power consumption by up to 40%.

Figures

Figures reproduced from arXiv: 2411.15399 by the authors.

Figure 1
Figure 1. Less-is-More considers three distinct tool representations (Search Levels): individual tools (Level 1), tool clusters (Level 2), or the entire tool set (Level 3), whose latent spaces are constructed offline. At runtime, given a query and without providing any tools, the LLM (Recommender) generates tool descriptions “ideal” for the task. Given the LLM-recommended tool embeddings, the Controller identifies the most re… view at source ↗
Figure 2
Figure 2. Performance comparison of our method (LiS) at varying [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Performance comparison of our method (LiS) at varying [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 3 Pith papers

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

  1. RankMap: Priority-Aware Multi-DNN Manager for Heterogeneous Embedded Devices

    cs.LG 2024-11 conditional novelty 6.0 of 10

    A priority-aware multi-DNN scheduler using Monte Carlo tree search and a learned throughput estimator reports higher throughput and no starvation on one embedded board.

  2. Multi-Agent Geospatial Copilots for Remote Sensing Workflows

    cs.LG 2025-01 conditional novelty 5.0 of 10

    A hybrid multi-agent orchestrator (composition plus iterative reassessment) reports 60.3% agentic correctness on generated remote sensing workflows, about 17 percentage points above the single-agent GeoLLM-Engine baseline.

  3. Enhancing Accuracy and Maintainability in Nuclear Plant Data Retrieval: A Function-Calling LLM Approach Over NL-to-SQL

    cs.CL 2025-06 conditional novelty 4.0 of 10

    A function-calling LLM that selects pre-approved SQL functions outperformed direct NL-to-SQL in human-evaluated correctness for nuclear plant data retrieval.

Reference graph

Works this paper leans on

45 extracted references · 16 canonical work pages · cited by 3 Pith papers

  1. [1]

    Gorilla: Large language model connected with massive apis,

    S. G. Patil, T. Zhang, X. Wang, and J. E. Gonzalez, “Gorilla: Large language model connected with massive apis,” arXiv preprint arXiv:2305.15334, 2023

    arXiv 2023

  2. [2]

    Pythia: An edge first agent for state prediction in high-dimensional environments,

    A. Karatzas and I. Anagnostopoulos, “Pythia: An edge first agent for state prediction in high-dimensional environments,” IEEE Embedded Systems Letters, 2024

    2024

  3. [3]

    Llm inference unveiled: Survey and roofline model insights,

    Z. Yuan, Y . Shang, Y . Zhou, Z. Dong, C. Xue, B. Wu, Z. Li, Q. Gu, Y . J. Lee, Y . Yan et al. , “Llm inference unveiled: Survey and roofline model insights,” arXiv preprint arXiv:2402.16363 , 2024

    arXiv 2024

  4. [4]

    Function calling documentation,

    OpenAI, “Function calling documentation,” https://platform.openai.com/ docs/guides/function-calling

    work page

  5. [5]

    ReAct: Synergizing reasoning and acting in language models,

    S. Yao, J. Zhao, D. Yu, N. Du, I. Shafran, K. Narasimhan, and Y . Cao, “ReAct: Synergizing reasoning and acting in language models,” in Inter- national Conference on Learning Representations (ICLR) , 2023

    work page 2023

  6. [6]

    Single-path nas: Designing hardware-efficient convnets in less than 4 hours,

    D. Stamoulis, R. Ding, D. Wang, D. Lymberopoulos, B. Priyantha, J. Liu, and D. Marculescu, “Single-path nas: Designing hardware-efficient convnets in less than 4 hours,” in Joint European Conference on Machine Learning and Knowledge Discovery in Databases . Springer, 2019, pp. 481–497

    work page 2019

  7. [7]

    Phi-3 technical report: A highly capable language model locally on your phone,

    M. Abdin et al., “Phi-3 technical report: A highly capable language model locally on your phone,” arXiv preprint arXiv:2404.14219 , 2024

    arXiv 2024

  8. [8]

    Function calling,

    HuggingFace, “Function calling,” https://huggingface.co/models?other= function+calling, 2024

    work page 2024

Show all 45 references
  1. [9]

    Tool support,

    Ollama, “Tool support,” https://ollama.com/blog/tool-support, 2024

    2024

  2. [10]

    Tinyagent: Function calling at the edge,

    L. E. Erdogan, N. Lee, S. Jha, S. Kim, R. Tabrizi, S. Moon, C. Hooper, G. Anumanchipalli, K. Keutzer, and A. Gholami, “Tinyagent: Function calling at the edge,” arXiv preprint arXiv:2409.00608 , 2024

    2024 arXiv

  3. [11]

    Berkeley function calling leaderboard,

    F. Yan, H. Mao, C. C.-J. Ji, T. Zhang, S. G. Patil, I. Stoica, and J. E. Gonzalez, “Berkeley function calling leaderboard,” https://gorilla.cs. berkeley.edu/blogs/8 berkeley function calling leaderboard.html, 2024

    2024

  4. [12]

    Geollm-engine: A realistic environ- ment for building geospatial copilots,

    S. Singh, M. Fore, and D. Stamoulis, “Geollm-engine: A realistic environ- ment for building geospatial copilots,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2024

    2024

  5. [13]

    Nvidia jetson agx orin series,

    L. S. Karumbunathan, “Nvidia jetson agx orin series,” 2022

    2022

  6. [14]

    Qlora: Efficient finetuning of quantized llms,

    T. Dettmers, A. Pagnoni, A. Holtzman, and L. Zettlemoyer, “Qlora: Efficient finetuning of quantized llms,” Advances in Neural Information Processing Systems, vol. 36, 2024

    2024

  7. [15]

    Llm-pruner: On the structural pruning of large language models,

    X. Ma, G. Fang, and X. Wang, “Llm-pruner: On the structural pruning of large language models,” Advances in neural information processing systems, vol. 36, pp. 21 702–21 720, 2023

    2023

  8. [16]

    Efficient memory management for large language model serving with pagedattention,

    W. Kwon, Z. Li, S. Zhuang, Y . Sheng, L. Zheng, C. H. Yu, J. Gonzalez, H. Zhang, and I. Stoica, “Efficient memory management for large language model serving with pagedattention,” in Proceedings of the 29th Symposium on Operating Systems Principles , 2023, pp. 611–626

    2023

  9. [17]

    MiniLLM: Knowledge distilla- tion of large language models,

    Y . Gu, L. Dong, F. Wei, and M. Huang, “MiniLLM: Knowledge distilla- tion of large language models,” in The Twelfth International Conference on Learning Representations , 2024

    2024

  10. [18]

    Llmlingua: Com- pressing prompts for accelerated inference of large language models,

    H. Jiang, Q. Wu, C.-Y . Lin, Y . Yang, and L. Qiu, “Llmlingua: Com- pressing prompts for accelerated inference of large language models,” in Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, 2023, pp. 13 358–13 376

    2023

  11. [19]

    Llm in a flash: Efficient large language model inference with limited memory,

    K. Alizadeh, I. Mirzadeh, D. Belenko, K. Khatamifard, M. Cho, C. C. Del Mundo, M. Rastegari, and M. Farajtabar, “Llm in a flash: Efficient large language model inference with limited memory,” arXiv preprint arXiv:2312.11514, 2023

    2023 arXiv

  12. [20]

    The breakthrough memory solutions for improved performance on llm inference,

    B. Kim, S. Cha et al., “The breakthrough memory solutions for improved performance on llm inference,” IEEE Micro, vol. 44, no. 3, pp. 40–48, 2024

    2024

  13. [21]

    Specinfer: Accelerating large language model serving with tree-based speculative inference and ver- ification,

    X. Miao, G. Oliaro, Z. Zhang, X. Cheng, Z. Wang, Z. Zhang, R. Y . Y . Wong, A. Zhu, L. Yang, X. Shi et al. , “Specinfer: Accelerating large language model serving with tree-based speculative inference and ver- ification,” in Proceedings of the 29th ACM International Conference...

    2024

  14. [22]

    Llm inference serv- ing: Survey of recent advances and opportunities,

    B. Li, Y . Jiang, V . Gadepally, and D. Tiwari, “Llm inference serv- ing: Survey of recent advances and opportunities,” arXiv preprint arXiv:2407.12391, 2024

    2024 arXiv

  15. [23]

    Edge-llm: Enabling efficient large language model adaptation on edge devices via unified compression and adaptive layer voting,

    Z. Yu, Z. Wang, Y . Li, R. Gao, X. Zhou, S. R. Bommu, Y . Zhao, and Y . Lin, “Edge-llm: Enabling efficient large language model adaptation on edge devices via unified compression and adaptive layer voting,” in Proceedings of the 61st ACM/IEEE Design Automation Conference, 2024...

    2024

  16. [24]

    Routerbench: A benchmark for multi-llm routing system,

    Q. J. Hu, J. Bieker, X. Li, N. Jiang, B. Keigwin, G. Ranganath, K. Keutzer, and S. K. Upadhyay, “Routerbench: A benchmark for multi-llm routing system,” arXiv preprint arXiv:2403.12031 , 2024

    2024 arXiv

  17. [25]

    Frugalgpt: How to use large language models while reducing cost and improving performance,

    L. Chen, M. Zaharia, and J. Zou, “Frugalgpt: How to use large language models while reducing cost and improving performance,” arXiv preprint arXiv:2305.05176, 2023

    2023 arXiv

  18. [26]

    Visualwebarena: Evaluating multimodal agents on realistic visual web tasks,

    J. Y . Koh, R. Lo, L. Jang, V . Duvvur, M. C. Lim, P.-Y . Huang, G. Neubig, S. Zhou, R. Salakhutdinov, and D. Fried, “Visualwebarena: Evaluating multimodal agents on realistic visual web tasks,” ACL, 2024

    2024

  19. [27]

    Llm-dcache: Improving tool- augmented llms with gpt-driven localized data caching,

    S. Singh, M. Fore, A. Karatzas, C. Lee, Y . Jian, L. Shangguan, F. Yu, I. Anagnostopoulos, and D. Stamoulis, “Llm-dcache: Improving tool- augmented llms with gpt-driven localized data caching,” arXiv preprint arXiv:2406.06799, 2024

    2024 arXiv

  20. [28]

    Memserve: Context caching for disaggregated llm serving with elastic memory pool,

    C. Hu, H. Huang, J. Hu, J. Xu, X. Chen, T. Xie, C. Wang, S. Wang, Y . Bao, N. Sun et al. , “Memserve: Context caching for disaggregated llm serving with elastic memory pool,” arXiv preprint arXiv:2406.17565, 2024

    2024 arXiv

  21. [29]

    ToolLLM: Facilitating large language models to master 16000+ real-world APIs,

    Y . Qin, S. Liang, Y . Ye, K. Zhu, L. Yan, Y . Lu, Y . Lin, X. Cong, X. Tang, B. Qian, S. Zhao, L. Hong, R. Tian, R. Xie, J. Zhou, M. Gerstein, dahai li, Z. Liu, and M. Sun, “ToolLLM: Facilitating large language models to master 16000+ real-world APIs,” in The Twelfth Internat...

    2024

  22. [30]

    Octopus: On-device language model for function calling of software apis,

    W. Chen, Z. Li, and M. Ma, “Octopus: On-device language model for function calling of software apis,”arXiv preprint arXiv:2404.01549, 2024

    2024 arXiv

  23. [31]

    An llm compiler for parallel function calling,

    S. Kim, S. Moon, R. Tabrizi, N. Lee, M. W. Mahoney, K. Keutzer, and A. Gholami, “An llm compiler for parallel function calling,” arXiv preprint arXiv:2312.04511, 2023

    2023 arXiv

  24. [32]

    An llm-tool compiler for fused parallel function calling,

    S. Singh, A. Karatzas, M. Fore, I. Anagnostopoulos, and D. Stamoulis, “An llm-tool compiler for fused parallel function calling,” arXiv preprint arXiv:2405.17438, 2024

    2024 arXiv

  25. [33]

    Balancing throughput and fair execution of multi-dnn workloads on heterogeneous embedded devices,

    A. Karatzas and I. Anagnostopoulos, “Balancing throughput and fair execution of multi-dnn workloads on heterogeneous embedded devices,” IEEE Transactions on Emerging Topics in Computing , 2024

    2024

  26. [34]

    Geckopt: Llm system efficiency via intent-based tool selection,

    M. Fore, S. Singh, and D. Stamoulis, “Geckopt: Llm system efficiency via intent-based tool selection,” in Proceedings of the Great Lakes Symposium on VLSI 2024 , 2024, pp. 353–354

    2024

  27. [35]

    Mapformer: Attention-based multi- dnn manager for throughout & power co-optimization on embedded devices,

    A. Karatzas and I. Anagnostopoulos, “Mapformer: Attention-based multi- dnn manager for throughout & power co-optimization on embedded devices,” in 2024 IEEE/ACM International Conference on Computer- Aided Design (ICCAD) . IEEE/ACM, 2024, pp. 1–9

    2024

  28. [36]

    Contextual retrieval,

    Anthropic, “Contextual retrieval,” https://www.anthropic.com/news/ contextual-retrieval, 2024

    2024

  29. [37]

    Mpnet: Masked and permuted pre-training for language understanding,

    K. Song, X. Tan, T. Qin, J. Lu, and T.-Y . Liu, “Mpnet: Masked and permuted pre-training for language understanding,” Advances in neural information processing systems , vol. 33, pp. 16 857–16 867, 2020

    2020

  30. [38]

    Toolqa: A dataset for llm question answering with external tools,

    Y . Zhuang, Y . Yu, K. Wang, H. Sun, and C. Zhang, “Toolqa: A dataset for llm question answering with external tools,” Advances in Neural Information Processing Systems , vol. 36, pp. 50 117–50 143, 2023

    2023

  31. [39]

    Agglomerative clustering,

    scikit learn, “Agglomerative clustering,” https://scikit-learn.org/stable/api/ index.html, 2024

    2024

  32. [40]

    The faiss library,

    M. Douze, A. Guzhva et al., “The faiss library,” 2024

    2024

  33. [41]

    Evaluating tool-augmented agents in remote sensing platforms,

    S. Singh, M. Fore, and D. Stamoulis, “Evaluating tool-augmented agents in remote sensing platforms,” arXiv preprint arXiv:2405.00709 , 2024

    2024 arXiv

  34. [42]

    Hermes-2-pro-llama-3-8b,

    NousResearch, “Hermes-2-pro-llama-3-8b,” https://huggingface.co/ NousResearch/Hermes-2-Pro-Llama-3-8B

  35. [43]

    The llama 3 herd of models,

    A. Dubey et al. , “The llama 3 herd of models,” arXiv preprint arXiv:2407.21783, 2024

    2024 arXiv

  36. [44]

    Mistral 7b,

    A. Jiang et al., “Mistral 7b,” arXiv preprint arXiv:2310.06825 , 2023

    2023 arXiv

  37. [45]

    Qwen2 technical report,

    A. Yang et al. , “Qwen2 technical report,” arXiv preprint arXiv:2407.10671, 2024

    2024 arXiv

Pith tools

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