Experts are all you need: A Composable Framework for Large Language Model Inference

• URL: http://arxiv.org/abs/2511.22955v1
• Date: Fri, 28 Nov 2025 08:00:16 GMT
• Title: Experts are all you need: A Composable Framework for Large Language Model Inference
• Authors: Shrihari Sridharan, Sourjya Roy, Anand Raghunathan, Kaushik Roy,
• Abstract summary: Large Language Models (LLMs) have achieved state-of-the-art accuracies in a variety of natural language processing (NLP) tasks.<n>MoEs overcome this bottleneck by decoupling model capacity from computation by only activating a subset of parameters or "experts"
• Score: 8.747592414164687
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Large Language Models (LLMs) have achieved state-of-the-art accuracies in a variety of natural language processing (NLP) tasks. However, this success comes at the cost of increased model sizes which leads to additional computational burden. Mixture of Experts (MoEs) overcome this bottleneck by decoupling model capacity from computation by only activating a subset of parameters or "experts". However, these models require joint pretraining of these experts along with the router and do not model multi-step reasoning. In contrast, multi-agent frameworks improve reasoning by decomposing complex problems into modular subtasks. However, these frameworks rely on sequential "plan--act--observe" loops, which introduce significant latency. Our work, Comp-LLM, addresses these challenges by introducing a composable inference framework that enables cross-expert collaboration via an explicit sub-query dependency graph. Comp-LLM consists of three components: (1) A Sub-query Generator that decomposes an input query, assigns each sub-query to an appropriate expert using embedding similarity, and constructs a dependency graph; (2) A Query Executor that processes nodes in the graph and identifies opportunities for parallelism based on dependencies and resource constraints; and (3) A Response Aggregator that synthesizes intermediate expert responses into a coherent final answer. Across several benchmarks, Comp-LLM achieves up to 11.01% accuracy improvement over monolithic LLMs of similar size, while offering 1.67x--3.56x reduction in model size with no significant degradation relative to the largest model in its family. Additionally, Comp-LLM provides 1.1x--1.7x latency improvement compared to sequential sub-query processing.

Related papers

• UniT: Unified Multimodal Chain-of-Thought Test-time Scaling [85.590774707406]
  Unified models can handle both multimodal understanding and generation within a single architecture, yet they typically operate in a single pass without iteratively refining their outputs.<n>We introduce UniT, a framework for multimodal test-time scaling that enables a single unified model to reason, verify, and refine across multiple rounds.
  arXiv  Detail & Related papers  (2026-02-12T18:59:49Z)
• Agent Primitives: Reusable Latent Building Blocks for Multi-Agent Systems [31.54866828797577]
  Existing multi-agent systems (MAS) can handle complex problems by enabling collaboration among multiple agents.<n>Most MAS communicate primarily through natural language, making them vulnerable to error accumulation and instability in long-context, multi-stage interactions.<n>We propose textbfAgent Primitives, a set of reusable latent building blocks for LLM-based MAS.
  arXiv  Detail & Related papers  (2026-02-03T16:17:53Z)
• One LLM to Train Them All: Multi-Task Learning Framework for Fact-Checking [7.856998585396422]
  Large language models (LLMs) are reshaping automated fact-checking (AFC) by enabling unified, end-to-end verification pipelines.<n>We propose textbfmulti-task learning (MTL) as a more efficient alternative that fine-tunes a single model to perform claim detection, evidence ranking, and stance detection jointly.
  arXiv  Detail & Related papers  (2026-01-16T13:44:25Z)
• Think Then Embed: Generative Context Improves Multimodal Embedding [51.76690812535934]
  We propose a Think-Then-Embed (TTE) framework for Universal Multimodal Embeddings (UME), composed of a reasoner and an embedder.<n>By leveraging a powerful MLLM reasoner, we achieve state-of-the-art performance on the MMEB-V2 benchmark, surpassing proprietary models trained on massive in-house datasets.
  arXiv  Detail & Related papers  (2025-10-06T16:53:56Z)
• MR$^2$-Bench: Going Beyond Matching to Reasoning in Multimodal Retrieval [86.35779264575154]
  Multimodal retrieval is becoming a crucial component of modern AI applications, yet its evaluation lags behind the demands of more realistic and challenging scenarios.<n>We introduce MR$2$-Bench, a reasoning-intensive benchmark for multimodal retrieval.
  arXiv  Detail & Related papers  (2025-09-30T15:09:14Z)
• A2R: An Asymmetric Two-Stage Reasoning Framework for Parallel Reasoning [57.727084580884075]
  Asymmetric Two-Stage Reasoning framework designed to bridge gap between a model's potential and its actual performance.<n>A2R-Efficient is a "small-to-big" variant that combines a Qwen3-4B explorer with a Qwen3-8B synthesizer.<n>Results show A2R is not only a performance-boosting framework but also an efficient and practical solution for real-world applications.
  arXiv  Detail & Related papers  (2025-09-26T08:27:03Z)
• SPARE: Single-Pass Annotation with Reference-Guided Evaluation for Automatic Process Supervision and Reward Modelling [58.05959902776133]
  We introduce Single-Pass.<n>with Reference-Guided Evaluation (SPARE), a novel structured framework that enables efficient per-step annotation.<n>We demonstrate SPARE's effectiveness across four diverse datasets spanning mathematical reasoning (GSM8K, MATH), multi-hop question answering (MuSiQue-Ans), and spatial reasoning (SpaRP)<n>On ProcessBench, SPARE demonstrates data-efficient out-of-distribution generalization, using only $sim$16% of training samples compared to human-labeled and other synthetically trained baselines.
  arXiv  Detail & Related papers  (2025-06-18T14:37:59Z)
• SUDER: Self-Improving Unified Large Multimodal Models for Understanding and Generation with Dual Self-Rewards [55.99492656542475]
  We propose textbfSUDER (textbfSelf-improving textbfUnified LMMs with textbfDual stextbfElf-textbfRewards), a framework reinforcing the understanding and generation capabilities of LMMs.
  arXiv  Detail & Related papers  (2025-06-09T17:38:45Z)
• Unifying Language Agent Algorithms with Graph-based Orchestration Engine for Reproducible Agent Research [32.92036657863354]
  Language agents powered by large language models (LLMs) have demonstrated remarkable capabilities in understanding, reasoning, and executing complex tasks.<n>However, developing robust agents presents significant challenges: substantial engineering overhead, lack of standardized components, and insufficient evaluation frameworks for fair comparison.<n>We introduce Agent Graph-based Orchestration for Reasoning and Assessment (AGORA), a flexible and abstraction framework that addresses these challenges.
  arXiv  Detail & Related papers  (2025-05-30T08:46:23Z)
• KG-QAGen: A Knowledge-Graph-Based Framework for Systematic Question Generation and Long-Context LLM Evaluation [3.618621510356872]
  KG-QAGen is a framework that extracts QA pairs at multiple complexity levels.<n>We construct a dataset of 20,139 QA pairs and open-source a part of it.<n>We evaluate 13 proprietary and open-source LLMs and observe that even the best-performing models are struggling with set-based comparisons.
  arXiv  Detail & Related papers  (2025-05-18T16:46:39Z)
• MAmmoTH-VL: Eliciting Multimodal Reasoning with Instruction Tuning at Scale [66.73529246309033]
  multimodal large language models (MLLMs) have shown significant potential in a broad range of multimodal tasks.<n>Existing instruction-tuning datasets only provide phrase-level answers without any intermediate rationales.<n>We introduce a scalable and cost-effective method to construct a large-scale multimodal instruction-tuning dataset with rich intermediate rationales.
  arXiv  Detail & Related papers  (2024-12-06T18:14:24Z)
• LaSagnA: Language-based Segmentation Assistant for Complex Queries [39.620806493454616]
  Large Language Models for Vision (vLLMs) generate detailed perceptual outcomes, including bounding boxes and masks. In this study, we acknowledge that the main cause of these problems is the insufficient complexity of training queries. We present three novel strategies to effectively handle the challenges arising from the direct integration of the proposed format.
  arXiv  Detail & Related papers  (2024-04-12T14:40:45Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.