datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
LDS-retrain-bank-adamw-wikitext2-N4656-bs8-seed1007
Retrain bank: WikiText-2 / GPT-2, random halves, seed 1007
This repository contains 100 fully retrained language models, not just scores.
Each model is GPT-2 (gpt2) fine-tuned on a different random 50% (2,328 documents)
of the 4,656-document WikiText-2 training set from
EleutherAI/bergson-wikitext-2-4656-chunks,
following the recipe of Bae et al. 2024, Training Data Attribution via Approximate
Unrolled Differentiation (App. B.1). retrained/base is trained on the full set with… See the full description on the dataset page: https://huggingface.co/datasets/EleutherAI/LDS-retrain-bank-adamw-wikitext2-N4656-bs8-seed1007.LDS-retrain-bank-adamw-wikitext2-N4656-bs8-seed1004
Retrain bank: WikiText-2 / GPT-2, random halves, seed 1004
This repository contains 100 fully retrained language models, not just scores.
Each model is GPT-2 (gpt2) fine-tuned on a different random 50% (2,328 documents)
of the 4,656-document WikiText-2 training set from
EleutherAI/bergson-wikitext-2-4656-chunks,
following the recipe of Bae et al. 2024, Training Data Attribution via Approximate
Unrolled Differentiation (App. B.1). retrained/base is trained on the full set with… See the full description on the dataset page: https://huggingface.co/datasets/EleutherAI/LDS-retrain-bank-adamw-wikitext2-N4656-bs8-seed1004.LDS-retrain-bank-adamw-wikitext2-N4656-bs8-seed1006
Retrain bank: WikiText-2 / GPT-2, random halves, seed 1006
This repository contains 100 fully retrained language models, not just scores.
Each model is GPT-2 (gpt2) fine-tuned on a different random 50% (2,328 documents)
of the 4,656-document WikiText-2 training set from
EleutherAI/bergson-wikitext-2-4656-chunks,
following the recipe of Bae et al. 2024, Training Data Attribution via Approximate
Unrolled Differentiation (App. B.1). retrained/base is trained on the full set with… See the full description on the dataset page: https://huggingface.co/datasets/EleutherAI/LDS-retrain-bank-adamw-wikitext2-N4656-bs8-seed1006.LDS-retrain-bank-adamw-wikitext2-N4656-bs8-seed1008
Retrain bank: WikiText-2 / GPT-2, random halves, seed 1008
This repository contains 100 fully retrained language models, not just scores.
Each model is GPT-2 (gpt2) fine-tuned on a different random 50% (2,328 documents)
of the 4,656-document WikiText-2 training set from
EleutherAI/bergson-wikitext-2-4656-chunks,
following the recipe of Bae et al. 2024, Training Data Attribution via Approximate
Unrolled Differentiation (App. B.1). retrained/base is trained on the full set with… See the full description on the dataset page: https://huggingface.co/datasets/EleutherAI/LDS-retrain-bank-adamw-wikitext2-N4656-bs8-seed1008.wikitext-103-raw-v1_gpt2-20k
Dataset Card for "wikitext-103-raw-v1_gpt2-20k"
More Information needed
LDS-retrain-bank-adamw-wikitext2-N4656-bs8-seed1005
Retrain bank: WikiText-2 / GPT-2, random halves, seed 1005
This repository contains 100 fully retrained language models, not just scores.
Each model is GPT-2 (gpt2) fine-tuned on a different random 50% (2,328 documents)
of the 4,656-document WikiText-2 training set from
EleutherAI/bergson-wikitext-2-4656-chunks,
following the recipe of Bae et al. 2024, Training Data Attribution via Approximate
Unrolled Differentiation (App. B.1). retrained/base is trained on the full set with… See the full description on the dataset page: https://huggingface.co/datasets/EleutherAI/LDS-retrain-bank-adamw-wikitext2-N4656-bs8-seed1005.bergson-wikitext-gpt2-leaderboard-bank
bergson leaderboard: retrain banks, scores and LDS/QLD results (WikiText GPT-2)
Everything behind the numbers on the bergson leaderboard,
for the model at EleutherAI/bergson-wikitext-gpt2-leaderboard.
path
what it is
bank/
the LDS ground truth: 100 random leave-1%-out subsets of the 4,608 training chunks (subsets.json) and each subset's measured loss change on the 50 test queries (validation.csv)
random/retrained/{base,subset_0..99}
the retrained models themselves… See the full description on the dataset page: https://huggingface.co/datasets/EleutherAI/bergson-wikitext-gpt2-leaderboard-bank.glm51-kld-reference-logits-wikitext-ctx2048-s512-20260517 ---
license: other
pretty_name: GLM-5.1 KLD Reference Logits WikiText ctx2048 s512
tags:
- logits
- kld
- glm-5.1
- vllm
- b12x
---
# GLM-5.1 KLD Reference Logits
Public cache of the reference logits used for GLM-5.1 NVFP4 / mixed
FP8_PB_WO KLD evaluation. These files are generated logits, not model
weights. They are stored as `logits_*.safetensors` with one tensor named
`logits`, shape `(2047, 154880)`, dtype `float32`.
##… See the full description on the dataset page: https://huggingface.co/datasets/festr2/glm51-kld-reference-logits-wikitext-ctx2048-s512-20260517.Stanza-Wikitext-2
Dataset Card for Stanza-Wikitext-2
Dataset Description
Stanza-Wikitext-2 is a structurally pristine, mathematically verified NLP dataset designed for multi-task language modeling, custom tokenizer training, structural NLP research, and mechanistic interpretability work.
It is a rigorously modernized and annotated derivative of the wikitext-2-raw-v1 corpus. Using the Stanford NLP Stanza neural pipeline, every token in the corpus has been explicitly mapped to its… See the full description on the dataset page: https://huggingface.co/datasets/EXOROBOURII/Stanza-Wikitext-2.qfs-smollm2-135m-wikitext2-native-v1
HF workflow d3dc69602aeb981f06bd9f4c726937f9
A root fidelity dataset in hidden form, produced by engines/tools/hf_capture.py from malaiwah/SmolLM2-135M-QFS-native-bf16.
The cut
the final hidden state handed to lm_head -- after the text model's final norm and immediately before the head matmul -- captured as the head module's input via torch.nn.Module.register_forward_pre_hook; replay applies the head ONLY (no final norm at replay time: the capture already sits… See the full description on the dataset page: https://huggingface.co/datasets/malaiwah/qfs-smollm2-135m-wikitext2-native-v1.wikitext__wikitext-103-raw-v1__clustering-1000wikitext-103-raw-pythia-word-tfidfwikitext-103-raw-pythia-word-tfidf-invfreq-topic-stratified-v1-articlesqfs-smollm2-135m-wikitext2-gptq-g32-v1
HF workflow 32c6ab05b0ceab1cecdceda838846388
A quant fidelity dataset in hidden form, produced by engines/tools/hf_capture.py from malaiwah/SmolLM2-135M-QFS-gptq-int4-g32.
The cut
the final hidden state handed to lm_head -- after the text model's final norm and immediately before the head matmul -- captured as the head module's input via torch.nn.Module.register_forward_pre_hook; replay applies the head ONLY (no final norm at replay time: the capture already sits… See the full description on the dataset page: https://huggingface.co/datasets/malaiwah/qfs-smollm2-135m-wikitext2-gptq-g32-v1.wikitext__wikitext-103-raw-v1__train__clustering-1000qfs-smollm2-135m-wikitext2-gptq-g64-v1
HF workflow 73f0a12a901c7368794a3a886f55b675
A quant fidelity dataset in hidden form, produced by engines/tools/hf_capture.py from malaiwah/SmolLM2-135M-QFS-gptq-int4-g64.
The cut
the final hidden state handed to lm_head -- after the text model's final norm and immediately before the head matmul -- captured as the head module's input via torch.nn.Module.register_forward_pre_hook; replay applies the head ONLY (no final norm at replay time: the capture already sits… See the full description on the dataset page: https://huggingface.co/datasets/malaiwah/qfs-smollm2-135m-wikitext2-gptq-g64-v1.wikitext__wikitext-103-raw-v1__clustering-3000wikitext-103-raw-pythia-word-tfidf-topic-stratified-v1-articlesflickr30k_wikitext-103-raw-clip_img_cap_ret_pairs-2022-09-26wikitext__wikitext-103-raw-v1__clustering-1000__median_cutoffqfs-smollm2-135m-wikitext2-rtn-g64-v1
HF workflow ee13c03b2f256042a17c9817e17a2f20
A quant fidelity dataset in hidden form, produced by engines/tools/hf_capture.py from malaiwah/SmolLM2-135M-QFS-rtn-int4-g64.
The cut
the final hidden state handed to lm_head -- after the text model's final norm and immediately before the head matmul -- captured as the head module's input via torch.nn.Module.register_forward_pre_hook; replay applies the head ONLY (no final norm at replay time: the capture already sits… See the full description on the dataset page: https://huggingface.co/datasets/malaiwah/qfs-smollm2-135m-wikitext2-rtn-g64-v1.wikitext__wikitext-103-raw-v1__clustering-1000__random_subsetwikitext-103-raw-pythia-tfidf-tokenlevelwikitext-103-raw-v2-tfidf-invfreq-topic-stratified-v1-articlesVALUE_wikitext2_got
Dataset Card for "VALUE_wikitext2_got"
More Information needed
flickr30k_wikitext-103-raw-clip_img_cap_ret_pairs-2022-09-25wikitext-103-raw-v4-df-tokenlevelwikitext-103-raw-pythia-tfidf-topic-stratified-v1-articleswikitext103_VALUE
Dataset Card for "wikitext103_VALUE"
More Information needed
wikitext-103-raw-v2-freq-invsqrt-topic-stratified-v1-articles