datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
Kimi-K3-CodexThis dataset was generated using teich by TeichAI
Kimi-K3 Codex traces
This directory contains raw agent trace files generated by teich.
JSONL files: 5
Model metadata: moonshotai/kimi-k3
Training-ready tools
Generated agent traces carry configured or recovered tool schemas so tools remain available for training even when a session did not call them.
Native Claude Code imports recover schemas for Claude Code and Claude Desktop built-ins, plus conservative… See the full description on the dataset page: https://huggingface.co/datasets/AletheiaResearch/Kimi-K3-Codex.kimi-k3-open-swe-distillation
Kimi K3 Open-SWE Distillation
Sanitized action-window data from a black-box Kimi K3 distillation experiment over NVIDIA Open-SWE-Traces. The rows retain native messages, exact tool schemas in tools, and metadata identifying the supervised assistant action.
Configurations
exact_replay: 69 action windows across 35 projects where moonshotai/kimi-k3, queried through OpenRouter, reproduced the source target action exactly.
critic_accepted: 198 source action windows… See the full description on the dataset page: https://huggingface.co/datasets/Infatoshi/kimi-k3-open-swe-distillation.thinkingcap-condensed-qwen3.8-glm5.2-kimi-k3
ThinkingCap Condensed — Qwen3.8 / GLM-5.2 / Kimi-K3
Condensed ThinkingCap-style reasoning traces for SFT.
1,985 traces: each row pairs a full multi-turn teacher trace (Qwen3.8-Max,
GLM-5.2 or Kimi K3, via
r0b0tlab/qwen3.8-max-glm5.2-kimi-k3-distillation)
with a condensed TC-style version (short <think> + definitive numbered
answer) generated by
bottlecapai/ThinkingCap-Qwen3.6-27B
using the thinkingcap system prompt.
Format: JSONL (data/condensed.jsonl), 1,985 rows, UTF-8.… See the full description on the dataset page: https://huggingface.co/datasets/Davd-b01/thinkingcap-condensed-qwen3.8-glm5.2-kimi-k3.kimi-k3-coding-and-debugging-traces
Kimi K3 Coding & Debugging Agent Traces
Generated by moonshiner — an open harness for
distilling verified, model-attested agentic coding traces.
Real, end-to-end agentic coding trajectories produced by
moonshotai/kimi-k3 driving the pi coding-agent runtime over
openrouter, at max reasoning. Each trajectory solves a concrete
repair or build task in a real repository — reading, editing, and running code
with tools — and is published only after its work verifiably passes —… See the full description on the dataset page: https://huggingface.co/datasets/gbeck/kimi-k3-coding-and-debugging-traces.kimi-k3-tiny-fidelity-root-v1
kimi-k3 random CPU fixture root
A root fidelity dataset in hidden form, produced by engines/tools/hf_capture.py from malaiwah/kimi-k3-tiny-random-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 after it). Same cut… See the full description on the dataset page: https://huggingface.co/datasets/malaiwah/kimi-k3-tiny-fidelity-root-v1.qwen3.8-max-glm5.2-kimi-k3-distillation-sua
qwen3.8-max-glm5.2-kimi-k3-distillation — System/User/Assistant format
Converted from r0b0tlab/qwen3.8-max-glm5.2-kimi-k3-distillation
(canonical config, current shard set train-*-of-00006; the stale of-00005 shards in the source repo were excluded).
Conversion date: 2026-08-20. License: inherited from the source — see LICENSE (controlled, noncommercial research scope).
Format
One JSON object per line, standard OpenAI-style chat format:
{"messages": [
{"role":… See the full description on the dataset page: https://huggingface.co/datasets/EuroswarmsInstitute/qwen3.8-max-glm5.2-kimi-k3-distillation-sua.kimi-k3-cyber-reasoning-distill
Kimi Cyber Reasoning
997 chain-of-thought records covering 13 cybersecurity disciplines and 4 systems engineering domains, distilled from the Kimi K3 reasoning model via API. Every record provides an explicit step-by-step <think> reasoning trace followed by a technical resolution, unified code diff fix, or structured tool invocation.
The dataset was curated as an anchor set for training, healing, and specializing compact reasoning models on systems security and tool calling… See the full description on the dataset page: https://huggingface.co/datasets/p-research/kimi-k3-cyber-reasoning-distill.SERA-KimiK3-Django-SWEAgent-Cliff32k-T1
SERA Kimi-K3 Django SWE-Agent — Cliff-chunked T1 (first rollout)
572 training records built from 210 Kimi-K3 SWE-agent trajectories on
Django, split to fit a 32,768-token context with
CliffCompaction instead of being truncated.
Why chunked
A 100+ step agent rollout does not fit a 32k training window — 76% of the
source T1 trajectories exceed it. Truncating them throws away most of the
supervision, and trains the model on a context format it never sees at… See the full description on the dataset page: https://huggingface.co/datasets/thientrangngv/SERA-KimiK3-Django-SWEAgent-Cliff32k-T1.SERA-KimiK3-Django-SWEAgent-Raw-T2
SERA Kimi-K3 Django SWE-Agent - RAW T2 (second rollout)
160 raw, pre-postprocess second-rollout agent trajectories generated with the
SERA SVG pipeline (paper),
using Kimi K3 as the teacher.
Each row is an independent attempt at the synthetic PR derived from a first rollout; target_patch
holds that first-rollout patch so soft verification can be recomputed at any r.
Companion: SERA-KimiK3-Django-SWEAgent-Raw-T1.
Schema
Mirrors allenai/Sera-*-T1/T2:
column… See the full description on the dataset page: https://huggingface.co/datasets/thientrangngv/SERA-KimiK3-Django-SWEAgent-Raw-T2.SERA-KimiK3-Django-SWEAgent-Raw-T1
SERA Kimi-K3 Django SWE-Agent - RAW T1 (first rollout)
300 raw, pre-postprocess first-rollout agent trajectories generated with the
SERA SVG pipeline (paper),
using Kimi K3 as the teacher.
Released raw so you can choose your own filtering, verification threshold and reasoning-trace handling.
Companion: SERA-KimiK3-Django-SWEAgent-Raw-T2.
Schema
Mirrors allenai/Sera-*-T1/T2:
column
notes
messages
JSON string - apply json.loads(). Raw SWE-agent history:… See the full description on the dataset page: https://huggingface.co/datasets/thientrangngv/SERA-KimiK3-Django-SWEAgent-Raw-T1.SERA-KimiK3-Django-SWEAgent-Cliff32k-T2
SERA Kimi-K3 Django SWE-Agent — Cliff-chunked T2 (second rollout)
227 training records built from 137 Kimi-K3 SWE-agent trajectories on
Django, split to fit a 32,768-token context with
CliffCompaction instead of being truncated.
Why chunked
A 100+ step agent rollout does not fit a 32k training window — 27% of the
source T2 trajectories exceed it. Truncating them throws away most of the
supervision, and trains the model on a context format it never sees at… See the full description on the dataset page: https://huggingface.co/datasets/thientrangngv/SERA-KimiK3-Django-SWEAgent-Cliff32k-T2.pi-kimi-k3-sft