schneiderkamplab/dfm13-multilingual-math-code-lb
dfm13_wave4_synthetic_lb_math_code 24107 complete conversations; 24107 native assistant targets. All user/tool history and tool definitions are preserved. Gemma native student rendering, thinking disabled. Generated and separately model-reviewed by Gemma4 26B A4B; automated judgments are fallible, not human or native-speaker certification. Includes unchanged original accepted conversations and narrowly recovered complete keep reviews rejected solely for an empty rationale.… See the full description on the dataset page: https://huggingface.co/datasets/schneiderkamplab/dfm13-multilingual-math-code-lb.
dfm13wave4syntheticlbmath_code
24107 complete conversations; 24107 native assistant targets. All user/tool history and tool definitions are preserved. Gemma native student rendering, thinking disabled. Generated and separately model-reviewed by Gemma4 26B A4B; automated judgments are fallible, not human or native-speaker certification.
Includes unchanged original accepted conversations and narrowly recovered complete keep reviews rejected solely for an empty rationale. Recovery revalidated saved raw calls, sources, untruncated renders and unique fingerprint ownership; no rationale was invented. No reject, repair, needs-verification or interrupted verdict was promoted.
See SOURCE_INVENTORY.json and row provenance. Source-specific attribution and terms remain applicable; this card makes no blanket license grant. HF upload is pending.
Attribution and source policy
See NOTICE.md and the per-conversation ATTRIBUTION.jsonl. Unknown OpenHermes component labels are disclosed under the identified upstream compilation and approved synthetic-source policy, not silently reassigned. This prepared card does not itself record an HF upload.