datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
metric-mamba-ml2021-hungyi-corpus
Dataset Card for "metric-mamba-ml2021-hungyi-corpus"
More Information needed
Vjepa_mamba_datasetdetails_CobraMamba__mamba-gpt-3b-v3
Dataset Card for Evaluation run of CobraMamba/mamba-gpt-3b-v3
Dataset Summary
Dataset automatically created during the evaluation run of model CobraMamba/mamba-gpt-3b-v3 on the Open LLM Leaderboard.
The dataset is composed of 64 configuration, each one coresponding to one of the evaluated task.
The dataset has been created from 2 run(s). Each run can be found as a specific split in each configuration, the split being named using the timestamp of the run.The "train" split… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard-old/details_CobraMamba__mamba-gpt-3b-v3.mamba-ml2021-hungyi-corpus
Dataset Card for "mamba-ml2021-hungyi-corpus"
More Information needed
details_CobraMamba__mamba-gpt-3b
Dataset Card for Evaluation run of CobraMamba/mamba-gpt-3b
Dataset Summary
Dataset automatically created during the evaluation run of model CobraMamba/mamba-gpt-3b on the Open LLM Leaderboard.
The dataset is composed of 64 configuration, each one coresponding to one of the evaluated task.
The dataset has been created from 2 run(s). Each run can be found as a specific split in each configuration, the split being named using the timestamp of the run.The "train" split is… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard-old/details_CobraMamba__mamba-gpt-3b.details_TRI-ML__mamba-7b-rw
Dataset Card for Evaluation run of TRI-ML/mamba-7b-rw
Dataset automatically created during the evaluation run of model TRI-ML/mamba-7b-rw on the Open LLM Leaderboard.
The dataset is composed of 62 configuration, each one coresponding to one of the evaluated task.
The dataset has been created from 2 run(s). Each run can be found as a specific split in each configuration, the split being named using the timestamp of the run.The "train" split is always pointing to the latest results.… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard-old/details_TRI-ML__mamba-7b-rw.fpga_cost_model_kernel_data_mamba_p2
FPGA HLS Kernel Cost-Model Data
Evolved Vitis HLS C++ kernels paired with their ground-truth Vitis HLS
csynth results. Each row is one generated program from an evolutionary FPGA
optimisation run, linked to its kernel source, evaluator report.json, and raw
synthesis report.
Each row carries a split label: train marks the original benchmarks used
to fit the analytical cost model's learned correction term, and holdout marks
benchmarks added afterwards that were not used for… See the full description on the dataset page: https://huggingface.co/datasets/adimnaku/fpga_cost_model_kernel_data_mamba_p2.snr-mamba-ml2021-hungyi-corpus
Dataset Card for "snr-mamba-ml2021-hungyi-corpus"
More Information needed
details_CobraMamba__mamba-gpt-3b-v2
Dataset Card for Evaluation run of CobraMamba/mamba-gpt-3b-v2
Dataset Summary
Dataset automatically created during the evaluation run of model CobraMamba/mamba-gpt-3b-v2 on the Open LLM Leaderboard.
The dataset is composed of 61 configuration, each one coresponding to one of the evaluated task.
The dataset has been created from 1 run(s). Each run can be found as a specific split in each configuration, the split being named using the timestamp of the run.The "train" split… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard-old/details_CobraMamba__mamba-gpt-3b-v2.repro-mamba-icl-outliers-actual
Actual Mamba ICL Outlier Reproduction Artifacts
Measured reproduction artifacts for ICML 2026 paper C41aLahRXZ, "How Can Mamba Learn In Context with Outliers and Generalize Provably?".
The scripts implement the paper synthetic setup: Eq. (3), Definitions 1-2, hinge-loss SGD, one-layer Mamba and linear Transformer control. Outputs in actual_outputs/ are from 5 seeds, 4000 SGD iterations, batch 128, and 4000 prompts per evaluation point.
HF GPU evidence:… See the full description on the dataset page: https://huggingface.co/datasets/Srishti280992/repro-mamba-icl-outliers-actual.InsectMambaDataU-Mamba2details_CobraMamba__mamba-gpt-7b-v1
Dataset Card for Evaluation run of CobraMamba/mamba-gpt-7b-v1
Dataset Summary
Dataset automatically created during the evaluation run of model CobraMamba/mamba-gpt-7b-v1 on the Open LLM Leaderboard.
The dataset is composed of 64 configuration, each one coresponding to one of the evaluated task.
The dataset has been created from 1 run(s). Each run can be found as a specific split in each configuration, the split being named using the timestamp of the run.The "train" split… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard-old/details_CobraMamba__mamba-gpt-7b-v1.Mamba2InLlama_0_875-best_of_n-PRM-3859129repro-mamba-icl-outliers
Reproduction: How Can Mamba Learn In Context with Outliers and Generalize Provably?
Independent reproduction of ICML 2026 paper C41aLahRXZ (arXiv 2510.00399,
titled on arXiv "Can Mamba Learn In Context with Outliers? A Theoretical
Generalization Analysis"), by Li, Lu, Cui, Chen, Wang.
No official code was released. This is a from-scratch implementation of the
paper's own theoretical model and synthetic ICL experiments.
What is implemented
mamba_icl.py — the… See the full description on the dataset page: https://huggingface.co/datasets/avithal/repro-mamba-icl-outliers.pretrain_dataset_mambarepro-sf-mamba
Reproduction: SF-Mamba: Rethinking State Space Model for Vision
Independent reproduction of ICML 2026 paper R9AUrEgZEq (arXiv 2603.16423)
by Yoshimura, Hayashi, Hoshino, Wang, Ohashi (Sony).
No official code or checkpoints are released ("We will release the source
code after publication"). SF-Mamba builds on the public
NVlabs/MambaVision backbone
(MambaVisionMixer(d_state=8, d_conv=3, expand=1), SSM on dim/2 channels).
What is reproduced
Claim
Approach… See the full description on the dataset page: https://huggingface.co/datasets/avithal/repro-sf-mamba.tiiuae__falcon-mamba-7b-details
Dataset Card for Evaluation run of tiiuae/falcon-mamba-7b
Dataset automatically created during the evaluation run of model tiiuae/falcon-mamba-7b
The dataset is composed of 38 configuration(s), each one corresponding to one of the evaluated task.
The dataset has been created from 1 run(s). Each run can be found as a specific split in each configuration, the split being named using the timestamp of the run.The "train" split is always pointing to the latest results.
An additional… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard/tiiuae__falcon-mamba-7b-details.SPScanner
Dataset
We release the training and evaluation dataset of Single-Pass Scanner. Our train set is mambaretriever_train.jsonl, our test set by categories is mambaretriever_test_per_category.json, and out test set by dataset is mambaretriever_test.json.
For more information about Single-Pass Scanner and the details of the datasets, check the Single-Pass Scanner Github
tiiuae__Falcon3-Mamba-7B-Base-details
Dataset Card for Evaluation run of tiiuae/Falcon3-Mamba-7B-Base
Dataset automatically created during the evaluation run of model tiiuae/Falcon3-Mamba-7B-Base
The dataset is composed of 38 configuration(s), each one corresponding to one of the evaluated task.
The dataset has been created from 1 run(s). Each run can be found as a specific split in each configuration, the split being named using the timestamp of the run.The "train" split is always pointing to the latest results.
An… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard/tiiuae__Falcon3-Mamba-7B-Base-details.MambaInLlama_0_50-best_of_n-PRM-3517996details_CobraMamba__mamba-gpt-7b-v2
Dataset Card for Evaluation run of CobraMamba/mamba-gpt-7b-v2
Dataset Summary
Dataset automatically created during the evaluation run of model CobraMamba/mamba-gpt-7b-v2 on the Open LLM Leaderboard.
The dataset is composed of 64 configuration, each one coresponding to one of the evaluated task.
The dataset has been created from 1 run(s). Each run can be found as a specific split in each configuration, the split being named using the timestamp of the run.The "train" split… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard-old/details_CobraMamba__mamba-gpt-7b-v2.train_data_imdb_simu
HH-RLHF-Helpful-Base Dataset
Summary
The HH-RLHF-Helpful-Base dataset is a processed version of Anthropic's HH-RLHF dataset, specifically curated to train models using the TRL library for preference learning and alignment tasks. It contains pairs of text samples, each labeled as either "chosen" or "rejected," based on human preferences regarding the helpfulness of the responses. This dataset enables models to learn human preferences in generating helpful responses… See the full description on the dataset page: https://huggingface.co/datasets/mamba413/train_data_imdb_simu.mamba-30m-datasetMamba2InLlama_0_875-best_of_n-PRM-3757864ptm_mamba_datasettiiuae__Falcon3-Mamba-7B-Instruct-details
Dataset Card for Evaluation run of tiiuae/Falcon3-Mamba-7B-Instruct
Dataset automatically created during the evaluation run of model tiiuae/Falcon3-Mamba-7B-Instruct
The dataset is composed of 38 configuration(s), each one corresponding to one of the evaluated task.
The dataset has been created from 1 run(s). Each run can be found as a specific split in each configuration, the split being named using the timestamp of the run.The "train" split is always pointing to the latest… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard/tiiuae__Falcon3-Mamba-7B-Instruct-details.Mamba2InLlama_0_50-best_of_n-PRM-3619944details_CobraMamba__mamba-gpt-3b-v4
Dataset Card for Evaluation run of CobraMamba/mamba-gpt-3b-v4
Dataset Summary
Dataset automatically created during the evaluation run of model CobraMamba/mamba-gpt-3b-v4 on the Open LLM Leaderboard.
The dataset is composed of 64 configuration, each one coresponding to one of the evaluated task.
The dataset has been created from 2 run(s). Each run can be found as a specific split in each configuration, the split being named using the timestamp of the run.The "train" split… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard-old/details_CobraMamba__mamba-gpt-3b-v4.micro-mamba-memory
MicroMamba
MicroMamba is a small-compute test of input-dependent state-space dynamics. Each
sequence contains distracting symbols, a few marked symbols, and a final query asking
for one marked item by ordinal position. Solving the task requires selective storage
and retrieval rather than ordinary next-token statistics.
The model uses a compact Mamba-inspired block with:
a causal depthwise convolution;
learned stable diagonal state dynamics;
input-dependent discretization, input… See the full description on the dataset page: https://huggingface.co/datasets/ARotting/micro-mamba-memory.