analiza aiqbench
This research introduces LogicIFGen and the LogicIFEval benchmark to rigorously evaluate Large Language Models on complex logic structures like conditions, loops, and function calls. The findings reveal a significant gap in current models' ability to follow intricate instructions, with most failing to execute more than 60% of verifiable tasks derived from code functions. This work highlights critical limitations in foundational instruction-following capabilities that hinder advanced reasoning and agentic behaviors.
LogicIF: Towards Complex Logic Instruction Following
arXiv:2508.09125v3 Announce Type: replace Abstract: Instruction following has catalyzed the recent era of Large Language Models (LLMs) and is the foundational skill underpinning more advanced capabilities such as reasoning and agentic behaviors. As tasks grow more challenging, the logic structures embedded in natural language instructions becomes increasingly intricate. However, how well LLMs perform on such logic-rich instructions remains under-explored. We propose LogicIFGen and LogicIFEval. LogicIFGen is a scalable, automated framework for generating verifiable instructions from code functions, which can naturally express rich logic such as conditions, loops, and function calls. We further curate a collection of complex code functions and use LogicIFGen to construct LogicIFEval, a benchmark comprising 426 verifiable logic-rich instructions. Our experiments demonstrate that current state-of-the-art LLMs still struggle to correctly follow the instructions in LogicIFEval. Most LLMs can only follow fewer than 60% of the instructions, revealing significant deficiencies in the instruction-following ability. Code and Benchmark: https://github.com/mianzhang/LogicIF
Abstract:Instruction following has catalyzed the recent era of Large Language Models (LLMs) and is the foundational skill underpinning more advanced capabilities such as reasoning and agentic behaviors. As tasks grow more challenging, the logic structures embedded in natural language instructions becomes increasingly intricate. However, how well LLMs perform on such logic-rich instructions remains under-explored. We propose LogicIFGen and LogicIFEval. LogicIFGen is a scalable, automated framework for generating verifiable instructions from code functions, which can naturally express rich logic such as conditions, loops, and function calls. We further curate a collection of complex code functions and use LogicIFGen to construct LogicIFEval, a benchmark comprising 426 verifiable logic-rich instructions. Our experiments demonstrate that current state-of-the-art LLMs still struggle to correctly follow the instructions in LogicIFEval. Most LLMs can only follow fewer than 60% of the instructions, revealing significant deficiencies in the instruction-following ability. Code and Benchmark: this https URL
Computation and Language (cs.CL); Machine Learning (cs.LG)
arXiv:2508.09125 [cs.CL]
(or arXiv:2508.09125v3 [cs.CL] for this version)
Text integral sub licența deschisă a publicației · 162 cuvinte
Subiecte conexe
[AINews] not much happened today
Latent Space · 18 sept. 2026
[AINews] not much happened today
Latent Space · 3 oct. 2026
[AINews] Reality Checks on AI News (Yegge shuts down Gas Town, Databricks’ +60% Astra cost)
Latent Space · 17 sept. 2026
Informe de riesgos de la IA de frontera (febrero–marzo de 2026)
METR · 19 mai 2026