karantonis/llada2-mini-uq-pickles
LLaDA 2.1 mini — UQ eval pickles (ue_manager_seed1) Durable backup of the LLaDA 2.1 mini uncertainty-quantification sweep pickles produced with lm-polygraph (branch feat/llada2-cache, April–May 2026). Each .pkl is a torch.load-able ue_manager dump containing per-sample stats (greedy_texts, target_texts, ...), estimations, gen_metrics, and metrics. Layout setup_a/ no-train baselines, reduced config (~34 est, K=10 sampling dropped) setup_b/ with-train baselines… See the full description on the dataset page: https://huggingface.co/datasets/karantonis/llada2-mini-uq-pickles.
Upload README.md with huggingface_hub
Delete setup_d/samsum_dfresh.pkl with huggingface_hub
Upload setup_d/samsum.pkl with huggingface_hub
Upload setup_b/triviaqa.pkl with huggingface_hub
Upload setup_d/xsum.pkl with huggingface_hub
Upload setup_a/wmt19_deen.pkl with huggingface_hub
Upload setup_d/triviaqa.pkl with huggingface_hub
Upload setup_a/xsum.pkl with huggingface_hub
Upload setup_c/mmlu.pkl with huggingface_hub
Upload setup_d/samsum_dfresh.pkl with huggingface_hub
Upload setup_d/gsm8k.pkl with huggingface_hub
Upload setup_d/coqa.pkl with huggingface_hub
Upload setup_a/wmt14_fren.pkl with huggingface_hub
Upload setup_d/truthfulqa.pkl with huggingface_hub
Upload setup_d/samsum.pkl with huggingface_hub
Upload setup_d/wmt19_deen.pkl with huggingface_hub
Upload setup_d/wmt14_fren.pkl with huggingface_hub
Upload setup_d/mmlu.pkl with huggingface_hub
Upload setup_a/samsum.pkl with huggingface_hub
Upload setup_a/triviaqa.pkl with huggingface_hub
Upload setup_a/mmlu.pkl with huggingface_hub
initial commit