DigiNews

Tech Watch Articles

← Back to articles

Can LLMs SAT? A practical look at LLM reasoning on SAT problems

Quality: 7/10 Relevance: 9/10

Summary

The article reports an informal study testing LLMs on SAT problems using CNF/DIMACS formats, comparing Gemini 3 Pro, GPT 5.2 Mini, and GPT 5.2. Results indicate that none of the models consistently reason, with performance deteriorating as problem size increases; the author concludes that LLMs cannot reliably perform formal reasoning and recommends external checks for critical tasks.

🚀 Service construit par Johan Denoyer