Pith. sign in

REVIEW 12 cited by

Are More LLM Calls All You Need? Towards Scaling Laws of Compound Inference Systems

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2403.02419 v2 pith:3WVIZI3Z submitted 2024-03-04 cs.LG cs.AIcs.CLcs.SYeess.SY

Are More LLM Calls All You Need? Towards Scaling Laws of Compound Inference Systems

classification cs.LG cs.AIcs.CLcs.SYeess.SY
keywords callsperformancenumbercompoundsystemsscalingvotefilter-vote
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
read the original abstract

Many recent state-of-the-art results in language tasks were achieved using compound systems that perform multiple Language Model (LM) calls and aggregate their responses. However, there is little understanding of how the number of LM calls - e.g., when asking the LM to answer each question multiple times and taking a majority vote - affects such a compound system's performance. In this paper, we initiate the study of scaling properties of compound inference systems. We analyze, theoretically and empirically, how the number of LM calls affects the performance of Vote and Filter-Vote, two of the simplest compound system designs, which aggregate LM responses via majority voting, optionally applying LM filters. We find, surprisingly, that across multiple language tasks, the performance of both Vote and Filter-Vote can first increase but then decrease as a function of the number of LM calls. Our theoretical results suggest that this non-monotonicity is due to the diversity of query difficulties within a task: more LM calls lead to higher performance on "easy" queries, but lower performance on "hard" queries, and non-monotone behavior can emerge when a task contains both types of queries. This insight then allows us to compute, from a small number of samples, the number of LM calls that maximizes system performance, and define an analytical scaling model for both systems. Experiments show that our scaling model can accurately predict the performance of Vote and Filter-Vote systems and thus find the optimal number of LM calls to make.

discussion (0)

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

Forward citations

Cited by 12 Pith papers

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

  1. Why Do Multi-Agent LLM Systems Fail?

    cs.AI 2025-03 unverdicted novelty 8.0

    The authors create the first large-scale dataset and taxonomy of failure modes in multi-agent LLM systems to explain their limited performance gains.

  2. Safety Under Scaffolding: How Evaluation Conditions Shape Measured Safety

    cs.SE 2026-03 conditional novelty 7.0

    Map-reduce scaffolding degrades measured safety mainly by stripping multiple-choice options (40–89% of the loss is format conversion); scaffold architecture explains only 0.4% of variance and composite safety scores h...

  3. RouterBench: A Benchmark for Multi-LLM Routing System

    cs.LG 2024-03 unverdicted novelty 7.0

    RouterBench supplies a standardized benchmark, 405k+ inference dataset, theoretical framework, and comparative analysis for multi-LLM routing systems.

  4. When More Sampling Hurts: The Modal Ceiling and Correlation Ceiling of Test-Time Scaling

    cs.LG 2026-06 unverdicted novelty 6.0

    Test-time sampling improves coverage but stalls at modal and correlation ceilings for answer selection, with the effective number of samples as the practical limit.

  5. DeepLook: Deeper Thinking with Lookahead

    cs.AI 2026-06 conditional novelty 6.0

    A training-free decoder that branches only at low-confidence reasoning segments and ranks continuations by average lookahead confidence reports equal or better accuracy at roughly 1/8 of a confidence-filtered baseline...

  6. Knowledge without Wisdom: Measuring Misalignment between LLMs and Intended Impact

    cs.LG 2026-03 unverdicted novelty 6.0

    Shared biases across LLMs from common pretraining misalign with teaching quality and negatively correlate with intended student learning outcomes, with model ensembles amplifying the misalignment.

  7. Large Language Monkeys: Scaling Inference Compute with Repeated Sampling

    cs.LG 2024-07 unverdicted novelty 6.0

    Repeated sampling scales problem coverage log-linearly with sample count, improving SWE-bench Lite performance from 15.9% to 56% using 250 samples.

  8. Blue Data Intelligence Layer: Streaming Data and Agents for Multi-source Multi-modal Data-Centric Applications

    cs.AI 2026-04 unverdicted novelty 5.0

    Blue DIL is a new architecture that unifies structured enterprise data, LLM world knowledge, and personal context through declarative query plans and agents for multi-source multi-modal applications.

  9. Scaling Teams or Scaling Time? Memory Enabled Lifelong Learning in LLM Multi-Agent Systems

    cs.MA 2026-03 unverdicted novelty 5.0

    LLMA-Mem improves long-horizon performance in LLM multi-agent systems over baselines while reducing cost and shows non-monotonic scaling where memory-enabled smaller teams can beat larger ones.

  10. A Scalable Multi-LLM Collaboration System with Retrieval-based Selection and Exploration-Exploitation-Driven Enhancement

    cs.CL 2025-07 unverdicted novelty 5.0

    SMCS coordinates 15 open-source LLMs via retrieval-based prior selection and exploration-exploitation posterior enhancement, outperforming GPT-4.1 by 5.36% and GPT-o3-mini by 5.28% on eight benchmarks.

  11. Latency and Cost of Multi-Agent Intelligent Tutoring at Scale

    cs.CY 2026-04 unverdicted novelty 3.0

    Priority PayGo keeps multi-agent tutoring responses under 4 seconds even at 50 concurrent users, while costs stay below textbook prices per student.

  12. Reinforcement Learning from Human Feedback

    cs.LG 2025-04 unverdicted novelty 2.0

    The book introduces the origins, mathematical setup, and optimization stages of RLHF including reward modeling, reinforcement learning, rejection sampling, and direct alignment algorithms.