Tool use and agents

Compute interpretation

Methods that allocate inference-time compute to tools, environments, retrieval, or action loops.

Supporting reading cards

• ReAct: Synergizing Reasoning and Acting in Language Models (2022, inference_time_compute_post_training)
• Program of Thoughts Prompting: Disentangling Computation from Reasoning for Numerical Reasoning Tasks (2022, inference_time_compute_post_training)
• Toolformer: Language Models Can Teach Themselves to Use Tools (2023, inference_time_compute_post_training)
• Voyager: An Open-Ended Embodied Agent with Large Language Models (2023, inference_time_compute_post_training)
• AlphaEvolve: A coding agent for scientific and algorithmic discovery (2025, search_simulation_science_compute)
• Kimi K2: Open Agentic Intelligence (2025, sparse_memory_efficient_scaling)
• DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models (2025, sparse_memory_efficient_scaling)
• Kimi K2.5: Visual Agentic Intelligence (2026, inference_time_compute_post_training)
• Qwen3.5-Omni Technical Report (2026, generative_media_compute)

Obsolete or less central under later compute

Track this only through linked reading cards; do not treat this method page as standalone evidence.