AIME-24

The AIME24 dataset/benchmark is used to evaluate the performance of Tool-Integrated Reasoning systems by providing a set of tasks that require strategic planning and self-correction through sequential tool invocation.

Papers using AIME-24 (5)

DeepTool: Scaling Interleaved Deliberation in Tool-Integrated Reasoning via Process-Supervised Reinforcement Learning2026What If We Allocate Test-Time Compute Adaptively?2026Team of Thoughts: Efficient Test-time Scaling of Agentic Systems through Orchestrated Tool Calling2026First Return, Entropy-Eliciting Explore2025LightTransfer: Your Long-Context LLM is Secretly a Hybrid Model with Effortless Adaptation2024