Commit History
17 days ago
-
Overlap check across v5's whole training line: MMLU flagged; TruthfulQA and HumanEval withdrawn
Verified signature45df242
19 days ago
-
Overlap check: withdraw HumanEval, flag TruthfulQA (v5 training data included their test questions)
Verified signaturev5 training data contains 50 of 164 HumanEval problems (all 30 scored) and 128 of 817 TruthfulQA questions. GSM8K, MMLU and HellaSwag checked clean (0 overlap).1e715b5
20 days ago
-
Correct benchmark results: standard lm-eval 0.4.10 scores; January spot check relabeled
Verified signatureReplaces the January 2026 spot-check figures (10 hand-written questions per subject) with lm-evaluation-harness 0.4.10 results on the full test sets. Results: https://huggingface.co/datasets/SparkSupernova/nova-mind-v5-lm-eval-results20d9200
-
Upload README.md with huggingface_hub
Verified signatureb865131 -
Add Nova Mind v5 model card with benchmark results
Verified signaturef2c6b45 -
Upload training_summary.json
Verified signature3bdd4bf -
Add Nova Mind v5 model card with benchmark results
Verified signature87dc49d -
Delete training_summary.json
Verified signaturefb9cbf4 -
Upload training_summary.json
Verified signature4660dd1 -
Upload vocab.json
Verified signaturefad7978 -
Upload merges.txt
Verified signaturef8fbd30 -
Upload chat_template.jinja
Verified signatureacfdde4 -
Upload added_tokens.json
Verified signature775e1da -
Upload special_tokens_map.json
Verified signaturea018de8 -
Upload tokenizer_config.json
Verified signature6716892 -
Upload tokenizer.json
Verified signature407bfd8 -
Upload adapter_model.safetensors
Verified signatureade50d1 -
Upload adapter_config.json
Verified signature5ab41ef -
Add Nova Mind v5 model card with benchmark results
Verified signature436540a -
initial commit
Verified signaturee83c4b0