Pack of LLMs: Model Fusion at Test-Time via Perplexity Optimization
Reviewed by Pithpith:BW2G3CEVopen to challenge →
read the original abstract
Fusing knowledge from multiple Large Language Models (LLMs) can combine their diverse strengths to achieve improved performance on a given task. However, current fusion approaches either rely on learning-based fusers that do not generalize to new LLMs, or do not take into account how well each LLM understands the input. In this work, we study LLM fusion at test-time, which enables leveraging knowledge from arbitrary user-specified LLMs during inference. We introduce Pack of LLMs (PackLLM), an effective method for test-time fusion that leverages each LLM's expertise, given an input prompt. PackLLM performs model fusion by solving an optimization problem for determining each LLM's importance, so that perplexity over the input prompt is minimized. First, our simple PackLLM-sim variant validates that perplexity is a good indicator for measuring each LLM's expertise. Second, our PackLLM-opt variant approximately solves the perplexity minimization problem via a greedy algorithm. The derived importance weights are used to combine the LLMs during inference. We conduct experiments with over 100 total LLMs on a diverse set of tasks. Experimental results show that (i) perplexity is a reliable measure for LLM fusion, (ii) PackLLM outperforms test-time fusion baselines by 1.89% accuracy points, and (iii) PackLLM can leverage new LLMs to improve performance over learning-based fusion approaches by 3.92-11.94% accuracy points.
This paper has not been read by Pith yet.
Forward citations
Cited by 6 Pith papers
-
DLLG: Dynamic Logit-Level Gating of LLM Experts
DLLG learns token-level fusion weights for LLM experts from sparse response supervision and outperforms routing, ensembling, and merging baselines on reasoning and code tasks.
-
Rethinking LLM Ensembling from the Perspective of Mixture Models
ME reinterprets LLM ensembling as a mixture model by sampling a single model stochastically at each token step, matching the ensemble distribution while invoking only one model per step for substantial speed gains.
-
Rethinking LLM Ensembling from the Perspective of Mixture Models
ME reinterprets LLM ensembling as token-level sampling from a mixture model, enabling single-model invocation per token with claimed mathematical equivalence to full ensembling and measured speedups of 1.78x-2.68x.
-
Less Diverse, Less Safe: The Indirect But Pervasive Risk of Test-Time Scaling in Large Language Models
Curtailing diversity in candidate pools for test-time scaling increases unsafe LLM outputs, as demonstrated by a reference-guided reduction protocol that evades standard safety classifiers across open and closed models.
-
A Scalable Multi-LLM Collaboration System with Retrieval-based Selection and Exploration-Exploitation-Driven Enhancement
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.
-
Harnessing Multiple Large Language Models: A Survey on LLM Ensemble
A systematic survey of LLM ensemble methods organized into a taxonomy of ensemble-before-inference, ensemble-during-inference, and ensemble-after-inference stages, with review of benchmarks, applications, and future d...
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.