Pages

DS-1000 Skyline LiveCodeBench LCB Skyline Other Domains Languages Thinking models Tool use Contamination OCaml Joint coding and testing JCT DS-1000 Probe geometry Fresh LCB Window Logprobs Reasoning traces Spider2 hardness TTP Explorer BFCL traces

Prompt analyses

Let's Think Through the Possibilities · Qwen3-8B · ProtoQA, HotPotQA, and TriviaQA prompt comparisons

Loading scored rollouts…