datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
Transmem_ecsd_minicpm5_1b_hotpotqa_n4_n8MiniCPM5-1B-atlas
juiceb0xc0de/MiniCPM5-1B-atlas
A brain atlas for openbmb/MiniCPM5-1B, a 1B on-device model with a 130k bilingual vocabulary. This is not a chat dataset or a benchmark. It is an internal-mechanics map, built by running activations through a corpus of prompts and scoring what each layer, component, head, and feature direction is doing.
If you want to know which parts of this model are safe to edit, where its output-vocabulary directions live, or which layers are carrying the most… See the full description on the dataset page: https://huggingface.co/datasets/juiceb0xc0de/MiniCPM5-1B-atlas.minicpm5-swe-native-eval-archive
MiniCPM5 原生 SWE 评测归档：32 run / 5842条任务记录
历史100-turn协议为11run/2200题，新600-turn协议为15run/3000题。各组独立目录；按同题配对，并保留调度和review差异。新协议表格见本文后半部分。
历史100-turn协议：11 run
11个run、2200题；同一固定100 Verified +100 Pro，各run终态与scratch/serving清理已核验。每题一个原始CC轨迹JSON，不重复保存每轮完整请求历史。
run
Verified 正确/已评分
V review
Pro 正确/已评分
P review
midtrain
51/93 (54.84%)
7
44/89 (49.44%)
11
step500
31/69 (44.93%)
31
21/60 (35.00%)
40
step1000
39/93 (41.94%)
7
19/83 (22.89%)
17
step1500
18/52… See the full description on the dataset page: https://huggingface.co/datasets/eigentom/minicpm5-swe-native-eval-archive.MiniCPM-RobotManip-LIBERO
MiniCPM-RobotManip LIBERO
This dataset contains the four LIBERO suites converted to LeRobot v3 format
for the MiniCPM-RobotManip LIBERO full-parameter fine-tuning example in
starVLA.
Dataset summary
Suite
Episodes
Frames
Videos
LIBERO-10
358
95,740
716
LIBERO-Goal
405
48,131
810
LIBERO-Object
450
66,294
900
LIBERO-Spatial
423
51,707
846
Total
1,636
261,872
3,272
Format: LeRobot v3
Frequency: 20 Hz
Cameras: agent view and wrist view
Video… See the full description on the dataset page: https://huggingface.co/datasets/openbmb/MiniCPM-RobotManip-LIBERO.latent-state-tracking-minicpm5
Tracking and Intervening on Latent State Dynamics in a Small Language Agent (MiniCPM5-2B)
Date: 2026-09-17
Model studied: openbmb/MiniCPM5-2B (2.52B params, 42 layers, hidden dim 2048)
Hardware: single RTX 3070 Ti (8GB) — all experiments run on consumer-grade hardware
Summary
We ask whether a small (2.5B-parameter) language model's hidden-state trajectory during generation contains a stable, low-dimensional structure that (a) is linearly decodable into task type… See the full description on the dataset page: https://huggingface.co/datasets/B2J/latent-state-tracking-minicpm5.minicpm5-sft3-3turn
minicpm5-sft3-3turn
Training / validation data for stage 1 (SFT) of the writing ladder behind
baiango/minicpm5-ul4b-story.
English fiction: each record is a templated writing instruction plus a
three-segment story distilled from teacher poolside/laguna-s-2.1
(3-turn chunked generation, automated quality gates).
Generation and training code for the full ladder: baiango/minicpm5-story-ladder.
Dataset summary
1,637 train / 40 valid records, one uniform schema across… See the full description on the dataset page: https://huggingface.co/datasets/baiango/minicpm5-sft3-3turn.minicpm5-sft-v10-deepseek-swe
nanocode-sft-ds-v10: cleaned DeepSeek-teacher SWE agent trajectories
Supervised fine-tuning rows for MiniCPM5-2B (tokenizer revision 0a45344e, chat template sha256 cc945752...), distilled
from a DeepSeek teacher (deepseek-v4.1-flash) acting in a Claude-Code-style harness (tools Bash, Read, Edit, Write) on
repository-level software-engineering tasks. Only resolved episodes (the final patch passed the task's tests) are used.
This is the DS component of the nanocode SFT mix v10; it… See the full description on the dataset page: https://huggingface.co/datasets/LingweiGu/minicpm5-sft-v10-deepseek-swe.minicpm5-brand-tools-eval-kit
MiniCPM5 brand-tools training and evaluation kit
Version 2.0 · 12 September 2026 · synthetic, offline evidence worlds
This kit is for the two existing functions verify_company_website and
find_customer_facing_pages. Their TypeScript implementations and function
schemas are copied unchanged from the prior brand-tools package. The kit creates
new synthetic environments and separates training targets from model-visible
evaluation prompts and private grader data.
Dataset… See the full description on the dataset page: https://huggingface.co/datasets/G33-k/minicpm5-brand-tools-eval-kit.minicpmv_overfit_lora
Model Card for Model ID
Model Details
Model Description
Developed by: [More Information Needed]
Funded by [optional]: [More Information Needed]
Shared by [optional]: [More Information Needed]
Model type: [More Information Needed]
Language(s) (NLP): [More Information Needed]
License: [More Information Needed]
Finetuned from model [optional]: [More Information Needed]
Model Sources [optional]
Repository: [More Information Needed]
Paper… See the full description on the dataset page: https://huggingface.co/datasets/cjfcsjt/minicpmv_overfit_lora.minicpm-v46-strict-cycle-runpod-serverless
MiniCPM Strict Caption Harness
Strict 9-field cycling harness for MiniCPM-V-4.6 v2 captions. The harness asks the model for one field at a time, cleans each field, and assembles the final caption with deterministic v2 headers.
Smoke test without loading the model:
cd /Users/dustinpainter/datasets/image-datasets
PYTHONPATH=tools/mini-cap-harness python3 tools/mini-cap-harness/run_strict_cycle.py \
--backend mock \
--input… See the full description on the dataset page: https://huggingface.co/datasets/jaddai/minicpm-v46-strict-cycle-runpod-serverless.minicpm5-2b-damage-labels
MiniCPM5-2B Damage Labels (MERNIK teacher)
Per-group measured quantization damage for MiniCPM5-2B (dense 2.6B, 42 layers).
What
damage_minicpm5_2b.jsonl — 169 rows: 1 BASELINE + 168 tied-group units.
Each unit row: the group dropped Q5_K → Q3_K while everything else stays at
Q5_K, scored by wikitext-2 PPL (-c 1024 -n 64 --seed 7).
{"unit": "ffn_down@7", "tensors": ["blk.7.ffn_down.weight"],
"ppl": 13.5364, "damage": 0.1732}
ssim_minicpm.npz — measured structural… See the full description on the dataset page: https://huggingface.co/datasets/wepiqx/minicpm5-2b-damage-labels.minicpm5-1b-quantization-benchmark
openbmb/MiniCPM5-1B 次世代量子化（Quanto FP8 / INT4 vs BNB 4bit）実測ベンチマークレポート
対象モデル: openbmb/MiniCPM5-1B (1.16B parameters, 128k context, LlamaForCausalLM)
検証ハードウェア: NVIDIA GeForce RTX 4070 Ti (12GB GDDR6X, Ada Lovelace, Compute Capability 8.9, 第4世代Tensor Core)
実行環境: Windows / Python 3.13 / PyTorch 2.6.0+cu124 / transformers 4.57.6 / optimum-quanto 0.2.7 / bitsandbytes 0.50.0
検証日: 2026-09-19 12:12:34
1. エグゼクティブサマリー（全体比較）
NVIDIA GeForce RTX 4070 Ti 実機環境において、標準ネイティブ… See the full description on the dataset page: https://huggingface.co/datasets/aoiandroid/minicpm5-1b-quantization-benchmark.FLAME-ReCap-YFCC15M-MiniCPM-Llama3-V-2_5
Dataset description
Recaptioned YFCC15M by MiniCPM-Llama3-V-2_5.
Uses
See https://github.com/MIV-XJTU/FLAME.
Citation
@article{cao2024flame,
title={FLAME: Frozen Large Language Models Enable Data-Efficient Language-Image Pre-training},
author={Cao, Anjia and Wei, Xing and Ma, Zhiheng},
journal={arXiv preprint arXiv:2411.11927},
year={2024}
}
@article{yao2024minicpmv,
title={MiniCPM-V: A GPT-4V Level MLLM on Your Phone},
author={Yao, Yuan… See the full description on the dataset page: https://huggingface.co/datasets/caj/FLAME-ReCap-YFCC15M-MiniCPM-Llama3-V-2_5.minicpm5-1b-SAEOne JumpReLU SAE per layer of MiniCPM5-1B. All 24 layers, complete.
MiniCPM5-1B: 24 layers, 1536-dim residual stream, 130,560-token bilingual vocab.
Every SAE in this repo: d_in=1536, 49,152 features (32x expansion), JumpReLU activation, streamed FineWeb-Edu, target sparsity L0=50. Same settings on every layer, no hyperparameter changes were applied in the run.
Each layer_NN_s0/ holds:
sae.pt - the weights
meta.json - config and final metrics
checkpoint_full.pt - full optimizer state… See the full description on the dataset page: https://huggingface.co/datasets/juiceb0xc0de/minicpm5-1b-SAE.minicpm-o45-native-gate-dataOfficial-native gate training/eval bundle for issue #8 (deployed gate_native.json, 8bq recipe, train_n=5228).
Contents (official_native_bundle.zip, 187 files, unpacks to official_native_bundle/data/):
Training tags caliboff, expoff, exp2off, exp3off, exp3zhoff, freshoff: frozen_native_<tag>_feats.shard*.npz (ids + 12288-d float32 X), frozen_native_<tag>_traces.jsonl.shard* (official-native answer_text, no_speak, eot_seen, n_ans_chunks), frozen_native_<tag>_judged.parquet (id, query… See the full description on the dataset page: https://huggingface.co/datasets/dyyfk/minicpm-o45-native-gate-data.minicpm5-sft-stage1-910c-pub
MiniCPM5-2B stage1 SFT tokenized packs (nanocode pipeline reproduction)
Sources: OpenBMB UltraData-SFT-2605 + UltraData-SFT-Agent-2609 (6 subsets, 6,278,234 records, used as published, no cleaning)
Preprocessing: MiniCPM5 pinned chat template, assistant-only loss mask, >128K overlapping windows (16K overlap, every target supervised once), packed to ~32K with per-record attention segments and position ids
Layout: HF save_to_disk format under stage1-tokenized/ (train 1,708,955… See the full description on the dataset page: https://huggingface.co/datasets/eigentom/minicpm5-sft-stage1-910c-pub.minicpm5-sft-swe-validation-200
MiniCPM5 SFT SWE Validation 200
A fixed, public task index for small-scale software-engineering evaluation. It contains two configurations, verified and pro, each with exactly 100 distinct tasks: 70 that the historical MiniCPM5-2B SFT baseline resolved and 30 that it did not resolve. Every row has benchmark, instance_id, task_id, sft_resolved, repo, and base_commit.
The task statements, repository contents, reference patches, and tests are not copied into this dataset. Join… See the full description on the dataset page: https://huggingface.co/datasets/eigentom/minicpm5-sft-swe-validation-200.minicpm5-stock-v2-forward-return
MiniCPM5 Stock v2 — Forward-Return Labels
Binary BUY/SELL stock-direction dataset where labels come from actual forward
5-day returns (BUY > +2%, SELL < -2%, middle band dropped), not news sentiment.
All features are strictly causal (no look-ahead): last 20 daily returns, RSI(14),
volume ratio vs 20d MA, 20d volatility, 5d/20d momentum, 20d relative strength vs SPY.
train_minicpm5_v2.jsonl — 5,056 rows, 16 tickers, class-balanced
val_minicpm5_v2.jsonl — 1,586 rows, 4 held-out… See the full description on the dataset page: https://huggingface.co/datasets/ewin-reg/minicpm5-stock-v2-forward-return.FLAME-ReCap-CC3M-MiniCPM-Llama3-V-2_5
Dataset description
Recaptioned CC3M by MiniCPM-Llama3-V-2_5.
Uses
The images are equivalent to https://huggingface.co/datasets/pixparse/cc3m-wds. Use data keys to index the original CC3M.
See https://github.com/MIV-XJTU/FLAME.
Citation
@article{cao2024flame,
title={FLAME: Frozen Large Language Models Enable Data-Efficient Language-Image Pre-training},
author={Cao, Anjia and Wei, Xing and Ma, Zhiheng},
journal={arXiv preprint arXiv:2411.11927}… See the full description on the dataset page: https://huggingface.co/datasets/caj/FLAME-ReCap-CC3M-MiniCPM-Llama3-V-2_5.MiniCPM5-2B-distill
MiniCPM5-2B-distill:
Distillation of MiniCPM5-2B (thinking off). Example:
{
"i":402, # i: unique ID
"kind":"chat", # kind of data; chat: user/assistant pairs, text: paragraphs
"cat":"explain", # category: can be used to sort data. E.g. "math", "greet", "creative", "text", "explain" etc...
"tok":65, # tokens amount: can be used to sort length. Most chat pairs are under 128 tokens and text under 512 tokens.
"text":"<|user|>\nWhat are the current… See the full description on the dataset page: https://huggingface.co/datasets/Dsg2/MiniCPM5-2B-distill.details_indischepartij__MiniCPM-3B-Hercules-v2.0
Dataset Card for Evaluation run of indischepartij/MiniCPM-3B-Hercules-v2.0
Dataset automatically created during the evaluation run of model indischepartij/MiniCPM-3B-Hercules-v2.0 on the Open LLM Leaderboard.
The dataset is composed of 63 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_indischepartij__MiniCPM-3B-Hercules-v2.0.details_indischepartij__MiniCPM-3B-OpenHermes-2.5-v2
Dataset Card for Evaluation run of indischepartij/MiniCPM-3B-OpenHermes-2.5-v2
Dataset automatically created during the evaluation run of model indischepartij/MiniCPM-3B-OpenHermes-2.5-v2 on the Open LLM Leaderboard.
The dataset is composed of 63 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… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard-old/details_indischepartij__MiniCPM-3B-OpenHermes-2.5-v2.minicpm5-gguf-stock-analyst
Stock Analyst Financial Trading Signals Dataset
A specialized dataset for fine-tuning Large Language Models (LLMs) to act as quantitative financial analysts. This dataset contains structured technical indicator data for stocks paired with their resulting directional trading signals (BUY, SELL, HOLD).
It is formatted specifically for Direct Preference Optimization (DPO) and Supervised Fine-Tuning (SFT), utilizing a hard-negative rejection strategy to force the model to learn… See the full description on the dataset page: https://huggingface.co/datasets/ewin-reg/minicpm5-gguf-stock-analyst.minicpm5-stage1-datadetails_indischepartij__MiniCPM-3B-Bacchus
Dataset Card for Evaluation run of indischepartij/MiniCPM-3B-Bacchus
Dataset automatically created during the evaluation run of model indischepartij/MiniCPM-3B-Bacchus on the Open LLM Leaderboard.
The dataset is composed of 63 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 is always… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard-old/details_indischepartij__MiniCPM-3B-Bacchus.wellness_voice_triplets_20250908_Chi_All_MiniCPM_r25to45details_openbmb__MiniCPM-2B-dpo-bf16-llama-format
Dataset Card for Evaluation run of openbmb/MiniCPM-2B-dpo-bf16-llama-format
Dataset automatically created during the evaluation run of model openbmb/MiniCPM-2B-dpo-bf16-llama-format on the Open LLM Leaderboard.
The dataset is composed of 63 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"… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard-old/details_openbmb__MiniCPM-2B-dpo-bf16-llama-format.minicpm5-vivamais-text-sft-v4
MiniCPM5 Viva Mais text SFT v4
This dataset contains the redacted training and evaluation artifacts used for
marinarosa/minicpm5-1b-vivamais-v4. It was built for Viva Mais, a local-first Portuguese WhatsApp
travel-agency copilot that answers grounded questions from an extracted CRM
context.
Files
data/train.jsonl: 4000 chat-format SFT rows.
data/eval/vivamais_qa_eval.jsonl: 158 dashboard QA eval
rows.
data/teacher/rio31_teacher_distill.jsonl: 80 accepted
rows… See the full description on the dataset page: https://huggingface.co/datasets/marinarosa/minicpm5-vivamais-text-sft-v4.details_gmonsoon__MiniCPM-2B-Base
Dataset Card for Evaluation run of gmonsoon/MiniCPM-2B-Base
Dataset automatically created during the evaluation run of model gmonsoon/MiniCPM-2B-Base on the Open LLM Leaderboard.
The dataset is composed of 63 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 is always pointing to the… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard-old/details_gmonsoon__MiniCPM-2B-Base.flashc-minicpm-sft