Jul 2026· Annual Conference on Genetic and Evolutionary Computation· pp. 271-279· 0 citations· 40 references
Computer Science
TL;DR
The method combines an interpretable cost-function formulation with a multi-objective evolutionary algorithm to discover Pareto-optimal ensemble configurations and introduces a two-stage combiner that produces diverse candidate responses and selects final outputs via embedding-based semantic consensus.
Abstract
Small Language Models (SLMs) offer efficient and practical alternatives to large-scale models in resource-constrained environments. We present a principled framework for constructing SLM ensembles that jointly optimize three competing objectives: prediction accuracy, output diversity, and fairness. Our method combines an interpretable cost-function formulation with a multi-objective evolutionary algorithm to discover Pareto-optimal ensemble configurations. We further introduce a two-stage combiner that produces diverse candidate responses and selects final outputs via embedding-based semantic consensus. Experiments on the MentalChat16k mental-health dialogue dataset show that the best-performing SLM ensemble configurations can match or surpass a fine-tuned Llama 3.1 70B model, achieving improvements of 0.86% in ROUGE-1, 5.84% in ROUGE-2, 4.93% in ROUGE-L, and 7.01% in semantic similarity. These results indicate that strategically orchestrated ensembles of small models can offer competitive or superior performance to significantly larger LLMs, while providing greater flexibility, interpretability, and accessibility for researchers operating under limited resources.
Aligning large language models (LLMs) with human values remains a major challenge, especially for trustworthy AI. While existing approaches such as RLHF, CAI, and their variants have achieved promising results, they often rely on a single-agent framework and a unified reward system. This limits their ability to capture ethical pluralism, adapt to diverse moral contexts, and reflect the dynamics of multi-agent moral reasoning. In this work, we propose a framework that utilizes multilayer combinatorial fusion for contextual value alignment (MCF-CVA). At the first layer of the framework, it instantiates multiple moral agents, each fine-tuned to represent a distinctive value. Their outputs are then expanded combinatorially using both score- and rank-combinations as well as average and weighted aggregations. These combined models are then reduced to the same number of initial moral agents. This expansion and reduction (EAR) process continues for multi-layers until a stopping criterion is reached. The MCF-CVA framework leverages cognitive diversity between agents to mitigate conflicts and redundancies across multiple agents, producing responses that better reflect contextual human values. The framework using the EAR algorithm is performed on the dual architecture of Euclidean score space and Kemeny rank space. Empirical evaluations demonstrated that the proposed framework outperforms single-agent baselines, multi-agent single-layer results, and previous aggregation approaches on standard metrics, showing that the MCF-CVA framework provides a robust and effective mechanism for advancing contextual value alignment in LLMs.
Yuanhong Wu, Djallel Bouneffouf, D. Hsu· 0 citations
This work proposes StrategyBench, which selects strategy-inducible tasks from BIG-Bench, constructs reference strategies, and defines evaluation metrics along two dimensions: strategy quality and downstream utility, and experiments show that explicit strategy utility differs substantially across task categories and depends on both strategy generation and execution conditions.
Jinghan Tan, Yuanzhe Wang, Lu Chen et al.· 0 citations
Parameter-Efficient Fine-Tuning (PEFT) commonly adapts large language models using a single shared Low-Rank Adapter (LoRA). This shared optimization space often suffers from interference when adapting heterogeneous task sequences, leading to poor transfer and catastrophic forgetting. Existing approaches mainly improve adapter expressiveness by increasing parameter capacity or composing multiple adapters, yet they still rely on a shared optimization path. In this paper, we propose an optimization-path organization framework for parameter-efficient fine-tuning of large language models, implemented as an automatic multi-policy PEFT architecture. Specifically, optimization-compatible adaptation paths are automatically organized through task grouping and task sequencing under a fixed parameter budget. The organized optimization paths are implemented as independent Quantized Low-Rank Adapters (QLoRA), enabling heterogeneous tasks to be optimized in decoupled adaptation spaces while preserving positive transfer among compatible tasks. Experiments on the TRACE benchmark demonstrate that performance consistently improves from conventional single-policy PEFT to multi-policy PEFT, with the proposed automatic multi-policy framework achieving the best performance of 44.78 under the same trainable capacity. This suggests that optimization-path organization is more effective than simply increasing adapter capacity for heterogeneous parameter-efficient fine-tuning.
Jiajia Tang, Sizhe Yuen, Francisco Gomez Medina et al.· 0 citations
Parameter-efficient fine-tuning (PEFT) methods have become essential for adapting large language models (LLMs) to downstream tasks without updating all parameters. However, the optimal PEFT strategy varies significantly across model architectures, task types, and resource constraints, making method selection a non-trivial problem. This paper presents PEFT-Assess, a systematic framework for evaluating the adaptability of four mainstream PEFT methods—LoRA, QLoRA, Adapter, and Prefix-Tuning—across five evaluation dimensions: accuracy, F1-score, training speed, memory efficiency, and generalization capability. We benchmark these methods on LLaMA-7B and Qwen-7B across six NLP tasks from the GLUE and SuperGLUE benchmarks. Furthermore, we propose a gradient boosting-based strategy selector that predicts the optimal PEFT method given task characteristics and resource budgets. Experiments show that LoRA achieves the best overall adaptability score of 0.847, while our strategy selector attains 91.2% prediction accuracy for recommending the optimal PEFT method.
Zhaoyang Hu, Shixuan Cao, Yuqi Sun et al.· International Conference on...· 0 citations
PPL-Factory is proposed, a simple and interpretable data selection framework that combines task-aware perplexity-based scores and data budget-aware selection criteria that outperforms other state-of-the-art data selection methods using only $1\%$ of the training set.
OptiMerge is proposed, a novel multi-objective evolutionary framework tailored for LLM model merging that discovers well-structured Pareto fronts and achieves improved performance trade-offs compared to base models and grid-search baselines.
Sri Srinivasa Raju Modampuri, Aojun Lu, Yanan Sun· Proceedings of the Genetic a...· 0 citations