Note The open reasoning flagship - trained 230+ models. 777 likes.

Note Math reasoning at scale with multiple solution traces. 682 likes.

Note Meta's 1.1M questions backtranslated from web text. Shows better scaling.

Note 306K problems from NVIDIA's reasoning initiative. 380 likes.

Note Medical domain reasoning with o1-style thinking. 967 likes.

Note 1st place winner - Reasoning Datasets Competition.

Note Medical reasoning grounded in knowledge graphs. Competition highlight.

Note Battlefield strategy reasoning - novel domain.

Note Symbolism and narrative structure from classic texts.

Note Multi-level legal explanations with human annotations.

Note Reasoning chains from academic papers. 33 likes.

Note Impossible choices inspired by Gödel - ethics/philosophy.

Note Historical what-ifs with DeepSeek-R1 reasoning.

Note VC investment reasoning - Best Curator Use Case.

Note Classifying what actually needs reasoning.

Note Generating reasoning questions from FineWeb.

Note 250K CoT examples distilled from DeepSeek-R1. 100 likes.