datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
cc-traces-weka-062126-256k
semianalysisai/cc-traces-weka-062126-256k
WekaTrace corpus derived from SemiAnalysis Claude Code proxy traces. Built 2026-06-21 17:49:45 UTC via utils/agentic/build_weka_hf_dataset.py.
Derived from semianalysisai/cc-traces-weka-062126 by applying the 256k per-request cap and preserving the surviving requests' relative timestamps.
Filters
Trace version: exactly v7
min Anthropic requests per session: 20
Claude Code CLI ≥ 2.1.139 (every row)
peak concurrent… See the full description on the dataset page: https://huggingface.co/datasets/semianalysisai/cc-traces-weka-062126-256k.cc-traces-weka-with-subagents-060826-256k
semianalysisai/cc-traces-weka-with-subagents-060826-256k
WekaTrace corpus derived from SemiAnalysis Claude Code proxy traces. Built 2026-06-08 20:48:46 UTC via utils/agentic/build_weka_hf_dataset.py.
Derived from semianalysisai/cc-traces-weka-with-subagents-060826 by applying the 256k per-request cap and reshifting the surviving timeline.
Filters
Trace version: exactly v6
min main-agent turns per session: 20
Claude Code CLI ≥ 2.1.139 (every row)
peak concurrent… See the full description on the dataset page: https://huggingface.co/datasets/semianalysisai/cc-traces-weka-with-subagents-060826-256k.cc-traces-weka-with-subagents-052726-256k
CC Traces — Weka, With Subagents, in+out ≤ 256k (May 27 2026, deduped)
A 256k-context simulation corpus derived from
cc-traces-weka-with-subagents-052726.
Every individual request (main-agent and sub-agent inner) where
in + out > 256,000 proxy-tokenizer tokens has been dropped. After filtering,
each trace's request timeline is reshifted using each surviving entry's
think_time so the gaps created by removed requests collapse cleanly.
Build note (2026-05-27 rebuild):… See the full description on the dataset page: https://huggingface.co/datasets/semianalysisai/cc-traces-weka-with-subagents-052726-256k.cc-traces-weka-061526-256k
semianalysisai/cc-traces-weka-061526-256k
WekaTrace corpus derived from SemiAnalysis Claude Code proxy traces. Built 2026-06-15 19:13:07 UTC via utils/agentic/build_weka_hf_dataset.py.
Derived from semianalysisai/cc-traces-weka-061526 by applying the 256k per-request cap and preserving the surviving requests' relative timestamps.
Filters
Trace version: exactly v7
min Anthropic requests per session: 20
Claude Code CLI ≥ 2.1.139 (every row)
peak concurrent… See the full description on the dataset page: https://huggingface.co/datasets/semianalysisai/cc-traces-weka-061526-256k.longmino_256k_filtered
longmino_filtered
本目录为 Dolma 3 Longmino Pool 数据经简单过滤后的分片，按子集分子文件夹存放。数据无需转换为 Parquet 即可配合 Hugging Face datasets 使用（见 scripts/load_longmino_example.py）。
数据来源与处理
上游数据集：allenai/dolma3_longmino_pool（Dolma 3 Longmino Pool，非混合后的训练 mix；详见该数据集说明）。
本目录：仅对上游数据做了简单过滤得到的副本文档集合，不是 Ai2 官方发行物，也不代表上游数据的完整或最新版本。
用途说明：以自用（个人研究/训练）为主，无面向公众的推广义务；若他人获取副本，仍应遵守下方许可与上游要求。
许可
遵循上游 Open Data Commons Attribution License v1.0（ODC-By） 及 dolma3_longmino_pool 页面中的许可与使用说明；引用与责任边界请以 Ai2 /… See the full description on the dataset page: https://huggingface.co/datasets/sagels/longmino_256k_filtered.cc-traces-weka-061326-256k
semianalysisai/cc-traces-weka-061326-256k
WekaTrace corpus derived from SemiAnalysis Claude Code proxy traces. Built 2026-06-13 19:01:03 UTC via utils/agentic/build_weka_hf_dataset.py.
Derived from semianalysisai/cc-traces-weka-061326 by applying the 256k per-request cap and reshifting the surviving timeline.
Filters
Trace version: exactly v7
min Anthropic requests per session: 20
Claude Code CLI ≥ 2.1.139 (every row)
peak concurrent sub-agent groups ≤ 10… See the full description on the dataset page: https://huggingface.co/datasets/semianalysisai/cc-traces-weka-061326-256k.cc-traces-weka-with-subagents-060226-256k
semianalysisai/cc-traces-weka-with-subagents-060226-256k
WekaTrace corpus derived from SemiAnalysis Claude Code proxy traces. Built 2026-06-02 15:19:33 UTC via utils/build_weka_hf_dataset.py.
Derived from semianalysisai/cc-traces-weka-with-subagents-060226 by applying the 256k per-request cap and reshifting the surviving timeline.
Filters
Trace version: exactly v6
min main-agent turns per session: 20
Claude Code CLI ≥ 2.1.139 (every row)
peak concurrent sub-agent… See the full description on the dataset page: https://huggingface.co/datasets/semianalysisai/cc-traces-weka-with-subagents-060226-256k.cc-traces-weka-with-subagents-060526-256k
semianalysisai/cc-traces-weka-with-subagents-060526-256k
WekaTrace corpus derived from SemiAnalysis Claude Code proxy traces. Built 2026-06-05 15:50:34 UTC via utils/agentic/build_weka_hf_dataset.py.
Derived from semianalysisai/cc-traces-weka-with-subagents-060526 by applying the 256k per-request cap and reshifting the surviving timeline.
Filters
Trace version: exactly v6
min main-agent turns per session: 20
Claude Code CLI ≥ 2.1.139 (every row)
peak concurrent… See the full description on the dataset page: https://huggingface.co/datasets/semianalysisai/cc-traces-weka-with-subagents-060526-256k.cc-traces-weka-with-subagents-060226-no-max-subagents-256k
semianalysisai/cc-traces-weka-with-subagents-060226-no-max-subagents-256k
WekaTrace corpus derived from SemiAnalysis Claude Code proxy traces. Built 2026-06-02 16:09:47 UTC via utils/build_weka_hf_dataset.py.
Derived from semianalysisai/cc-traces-weka-with-subagents-060226-no-max-subagents by applying the 256k per-request cap and reshifting the surviving timeline.
Filters
Trace version: exactly v6
min main-agent turns per session: 20
Claude Code CLI ≥ 2.1.139… See the full description on the dataset page: https://huggingface.co/datasets/semianalysisai/cc-traces-weka-with-subagents-060226-no-max-subagents-256k.winglian__llama-3-8b-256k-PoSE-details
Dataset Card for Evaluation run of winglian/llama-3-8b-256k-PoSE
Dataset automatically created during the evaluation run of model winglian/llama-3-8b-256k-PoSE
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/winglian__llama-3-8b-256k-PoSE-details.