analiza aiqbench
MIRAGE introduces a novel inference-time framework that enhances LLM performance on complex reasoning tasks by dynamically switching between multiple conceptual perspectives, such as algebraic and probabilistic approaches. By integrating a selector and reasoner guided by reinforcement learning, the method significantly boosts accuracy on mathematical and logical benchmarks with minimal computational overhead compared to standard prompting techniques. This approach offers a scalable pathway for improving foundation models in scientific and analytical domains without requiring extensive retraining.
MIRAGE: Multi-Perspective Creative Language Model Reasoning with Reinforcement Learning Guidance
arXiv:2609.21554v1 Announce Type: new Abstract: Recent advances in Large Language Models (LLMs) have revolutionized artificial intelligence and how human interact with AIs. Despite impressive advancements, LLMs struggle with complex mathematical, scientific, and logical tasks. Inspired by human cognitive flexibility - our ability to dynamically switch mental perspectives - we propose MIRAGE (Multi-perspective Inference-time Reasoning via Agent-Guided Exploration), a novel inference-time creative thinking framework. MIRAGE includes a Selector that prioritizes effective conceptual perspectives (e.g., algebraic, probabilistic) and a Reasoner that sequentially solves tasks until a confident solution emerges, otherwise aggregating multiple perspectives. Tested on GSM8K, MATH500, MMLU-Pro, and Game-of-24 benchmarks, MIRAGE consistently outperforms methods like Chain-of-Thought and diverse prompting ensembles, significantly boosting accuracy with minimal inference overhead, providing a scalable solution for practical applications.
Abstract:Recent advances in Large Language Models (LLMs) have revolutionized artificial intelligence and how human interact with AIs. Despite impressive advancements, LLMs struggle with complex mathematical, scientific, and logical tasks. Inspired by human cognitive flexibility - our ability to dynamically switch mental perspectives - we propose MIRAGE (Multi-perspective Inference-time Reasoning via Agent-Guided Exploration), a novel inference-time creative thinking framework. MIRAGE includes a Selector that prioritizes effective conceptual perspectives (e.g., algebraic, probabilistic) and a Reasoner that sequentially solves tasks until a confident solution emerges, otherwise aggregating multiple perspectives. Tested on GSM8K, MATH500, MMLU-Pro, and Game-of-24 benchmarks, MIRAGE consistently outperforms methods like Chain-of-Thought and diverse prompting ensembles, significantly boosting accuracy with minimal inference overhead, providing a scalable solution for practical applications.
18 pages, 5 figures. Accepted at the ICML 2025 Workshop on Multi-Agent Systems in the Era of Foundation Models: Opportunities, Challenges and Futures (MAS-2025)
Computation and Language (cs.CL)
arXiv:2609.21554 [cs.CL]
(or arXiv:2609.21554v1 [cs.CL] for this version)
https://doi.org/10.48550/arXiv.2609.21554 arXiv-issued DOI via DataCite (pending registration)
Text integral sub licența deschisă a publicației · 161 cuvinte
