GLM-5.2-Logic-Puzzles

<p align="center"> <img alt="arc" src="https://huggingface.co/datasets/ianncity/GLM-5.2-Logic-Puzzles/resolve/main/glm-1.webp" width="200"> </p> <div align="center"> ## <span style="color:#EDEADE">GLM-5.2 · Logical Puzzles</span> **6000x traces distilled from `GLM-5.2` on `High` reasoning** ### <span style="color:#EDEADE">Token Count</span>: <code>5M~?</code> </div> --- ## Distribution: ``` Puzzles: •Tokenization blindless ex: counting the r's in strawberry •Goal reasoning ex: the car wash test (theres no car wash question exactly just prompts like it so its not just benchmaxxing) •Reading comprehension traps •Temporal reasoning •Many other categories not worth mentioning Prompts generated with Diffusion Gemma 27B answered by GLM-5.2, Then the outputs are near deduplicated then quality pruned by a judge model, so a lot of completions went to waste here Also my smallest dataset ever, turns out theres not many different types of logical puzzles ``` --- <p align="center" style="font-size:11px;opacity:0.7"> You can use this dataset for any purpose and you dont need to credit me, preferably dont claim it as your own. </p> <br> <div align="left"> <img src="https://cdn-uploads.huggingface.co/production/uploads/6909e3efe7497f3dbbe640ba/AT1xJ2uYHT7LjFsxewu-U.png" width="140"> </div> #### hi - ianncity <img src="https://preview.redd.it/steam-happy-but-high-quality-v0-22ku6htw4u0c1.png?width=640&crop=smart&auto=webp&s=221735cb09dc3d4c1c7349e3187e752f6fe775e4" width="20" height="20" style="display:inline-block;vertical-align:middle;margin:0 3px;">