datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
GLM-5.2-AgentThis dataset was generated using teich by TeichAI
GLM-5.2 Agent traces
This directory contains raw agent trace files generated by teich.
JSONL files: 319
Model metadata: glm-5.2
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 name-derived MCP… See the full description on the dataset page: https://huggingface.co/datasets/AletheiaResearch/GLM-5.2-Agent.GLM-5.1-Reasoning-1M-Cleaned
GLM-5.1-Reasoning-1M-Cleaned
GLM-5.1-Reasoning-1M-Cleaned is a cleaned and reformatted derivative of Kassadin88/GLM-5.1-1000000x. It preserves the original four-subset layout (main, PHD-Science, Multilingual-STEM, Math) while converting every example into a unified SFT-ready schema with explicit conversations, input, output, domain, and meta fields.
This release was prepared from the original dataset published by Kassadin88.
Summary
Teacher model in the data: GLM-5.1… See the full description on the dataset page: https://huggingface.co/datasets/Jackrong/GLM-5.1-Reasoning-1M-Cleaned.GLM-5.3-Flash-calibration-activations-v1
GLM-5.3-Flash calibration activations v1 (BF16, natural routing)
Per-layer block-input activations of zai-org/GLM-5.3-Flash-BF16 @ b1967181 over 92x2048
tokens of the exllamav3 standard_cal_data corpus (pinned): per context, layer_NNN.attn_in
and layer_NNN.mlp_in (bf16, post-norm linear inputs; mlp_in is the router + expert gate/up
input) and layer_NNN.router_logits (fp32, natural top-8 routing ground truth).
Per-expert Hessians E[xx^T], routing statistics and down-proj inputs… See the full description on the dataset page: https://huggingface.co/datasets/malaiwah/GLM-5.3-Flash-calibration-activations-v1.GLM-5.2-Conversation
GLM-5.2 · Conversation-50000x
50,000x traces distilled from GLM-5.2 on High reasoning
Token Count: 120M
Distribution:
Speaking domains:
•Greetings
•Customer Support
•Step by step explanations
•Motivational language
•Logical Questions
•Creative Writing
STEM:
•Algebra, calculus, quantum mechanics concepts
•Astromony and astrophysics
•Datascience and machine learning
•Biology
Programming:… See the full description on the dataset page: https://huggingface.co/datasets/ianncity/GLM-5.2-Conversation.GLM-5.2-Finance-80000x
GLM-5.2 · Finance-80000x
80,000x financial related traces distilled from GLM-5.2 on High reasoning
Risk · Markets · Investments · Corporate Finance · Wealth Management
Token Count: 220M
Unique prompts generated with diffusion Gemma-27B answered by GLM-5.2
You can use this dataset for any purpose and you dont need to credit me, preferably dont claim it as your own.
hi - ianncity
glm52-demolition-data
GLM-5.2-Demolition — Training & Calibration Data
Apple Silicon AI hub ·
Model release ·
MLX code sample
Preview scope, checked September 10, 2026: the default Hub viewer indexes
87,586 rows (84,231 train, 3,277 validation, 78 test). The original release
total below describes the broader JSONL repository. Use the file browser and
explicit file selections when reusing a particular corpus. The hub includes
a checked download example for the seven-row MLX code sample.
The data… See the full description on the dataset page: https://huggingface.co/datasets/philipjohnbasile/glm52-demolition-data.glm53-fidelity-exl3-wrld-k4-v1
fidelity--glm53.malaiwah.quant.exl3-wrld-k4
A quant fidelity dataset in hidden form, produced by engines/tools/hf_capture.py from wrldsuksgo2mars/GLM-5.3-EXL3-K4-v1.
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… See the full description on the dataset page: https://huggingface.co/datasets/malaiwah/glm53-fidelity-exl3-wrld-k4-v1.glm53-flash-fidelity-root-v1
fidelity--glm53flash.malaiwah.root.bf16
A root fidelity dataset in hidden form, produced by engines/tools/hf_capture.py from zai-org/GLM-5.3-Flash-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/glm53-flash-fidelity-root-v1.GLM-5.1-Reasoning-Main-Sharded
GLM-5.1 reasoning main — sequential shards
Byte-preserving 100 MB JSONL shards of the main subset from Jackrong/GLM-5.1-Reasoning-1M-Cleaned, by Jackrong, derived upstream from Kassadin88/GLM-5.1-1000000x. All credit for the original data and cleaning belongs to those publishers.
Only main.jsonl is included. No filtering, shuffling, schema changes, tokenization or truncation. Original JSON fields and complete records are preserved. Shards retain upstream order; random shard… See the full description on the dataset page: https://huggingface.co/datasets/placeholderlabs/GLM-5.1-Reasoning-Main-Sharded.glm52-fidelity-nvfp4-nvidia-v1
fidelity--glm52.malaiwah.quant.nvfp4-nvidia
A quant fidelity dataset in hidden form, produced by engines/tools/hf_capture.py from nvidia/GLM-5.2-NVFP4.
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/glm52-fidelity-nvfp4-nvidia-v1.GLM-5.1-1000000x
GLM-5.1-1000000x
1,003,589 reasoning traces distilled by GLM-5.1, using questions from KIMI-K2.5-1000000x.
Each entry contains a full chain-of-thought reasoning trace followed by the final answer, generated by GLM-5.1.
Complete! All 1,003,589 prompts distilled successfully.
████████████████████████████████ 100%
Data Distribution
Subset
Count
Proportion
Est. Tokens
Domain
main
598,366
59.6%
~3.04B
General reasoning & instruction-following
Math… See the full description on the dataset page: https://huggingface.co/datasets/clzoro/GLM-5.1-1000000x.ale-bench-glm5.2-trajectories
ALE-Bench trajectories — glm5.2
CLI-agent trajectories on ALE-Bench
(Sakana AI; AtCoder Heuristic Contest long-horizon algorithm engineering),
problems ahc026 (Stack of Boxes) + ahc039 (Purse Seine Fishing),
lite version (5 public / ~150 private cases), 90-min wall-clock budget per task.
Agent: glm-5.2 requested via claude-code CLI on z.ai — NOTE: z.ai silently served glm-5.3 (per-response model field, 548/550 responses).
Harness (custom, not the paper's): agent works in… See the full description on the dataset page: https://huggingface.co/datasets/AgentNativeResearchLab/ale-bench-glm5.2-trajectories.GLM-5.2-Logic-Puzzles
GLM-5.2 · Logical Puzzles
6000x traces distilled from GLM-5.2 on High reasoning
Token Count: 5M~?
Distribution:
Puzzles:
•Tokenization blindless ex: counting the r's in strawberry
•Goal reasoning ex: the car wash test (theres no car wash question exactly just prompts like it so its not just benchmaxxing)
•Reading comprehension traps
•Temporal reasoning
•Many other categories not worth mentioning
Prompts… See the full description on the dataset page: https://huggingface.co/datasets/ianncity/GLM-5.2-Logic-Puzzles.GLM-5.2-FP8-nemotron-codealpaca
GLM-5.2-FP8-nemotron-codealpaca
Training data for UCloud-org/GLM-5.2-FP8-DFlash,
a DFlash speculative-decoding drafter for
zai-org/GLM-5.2-FP8.
A mix of code / math / chat prompts from two public instruction datasets
(see Composition); all assistant responses are regenerated by GLM-5.2-FP8 so the targets match the
verifier's own output distribution — the data recipe specified in the
DFlash paper (Appendix A.1).
800,022 single-turn conversations, English-dominant
Generation:… See the full description on the dataset page: https://huggingface.co/datasets/JessieWei/GLM-5.2-FP8-nemotron-codealpaca.glm53-flash-fidelity-exl3-tr3-6bpw-v1
fidelity--glm53flash.malaiwah.quant.tr3-6bpw
A quant fidelity dataset in hidden form, produced by engines/tools/hf_capture.py from malaiwah/GLM-5.3-Flash-TR3-6bpw.
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… See the full description on the dataset page: https://huggingface.co/datasets/malaiwah/glm53-flash-fidelity-exl3-tr3-6bpw-v1.glm-5.2-nvfp4-agentic-traces
GLM-5.2 NVFP4 agentic software traces
This snapshot contains 1,989 completed Verifiers invocation records generated
with RedHatAI/GLM-5.2-NVFP4-FP8.
manifest.jsonl is a compact index for filtering and inspection.
data/<arm>.jsonl contains the exact full graph records emitted by Verifiers.
configs/<arm>.toml contains the resolved configuration for each arm.
The snapshot retains successes, failures, truncations, and scoring metadata.
Use solved, reward, has_error, failure_labels… See the full description on the dataset page: https://huggingface.co/datasets/synquid/glm-5.2-nvfp4-agentic-traces.GLM-5.2-Science
GLM-5.2 · Science-50000x
50,000x traces distilled from GLM-5.2 on High reasoning
Physics · Chemistry · Biology
Token Count: 160M
Theres prompt overlap with my Kimi K2.5 dataset science subset, which I think those prompts are getting used in alot of places now
You can use this dataset for any purpose and you dont need to credit me, preferably dont claim it as your own.
hi - ianncity
glm53-flash-fidelity-fp8-v1
fidelity--glm53flash.malaiwah.quant.official-fp8
A quant fidelity dataset in hidden form, produced by engines/tools/hf_capture.py from zai-org/GLM-5.3-Flash.
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).… See the full description on the dataset page: https://huggingface.co/datasets/malaiwah/glm53-flash-fidelity-fp8-v1.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.GLM-5.1-Reasoning-1M-Cleaned
GLM-5.1-Reasoning-1M-Cleaned
GLM-5.1-Reasoning-1M-Cleaned is a cleaned and reformatted derivative of Kassadin88/GLM-5.1-1000000x. It preserves the original four-subset layout (main, PHD-Science, Multilingual-STEM, Math) while converting every example into a unified SFT-ready schema with explicit conversations, input, output, domain, and meta fields.
This release was prepared from the original dataset published by Kassadin88.
Summary
Teacher model in the data: GLM-5.1… See the full description on the dataset page: https://huggingface.co/datasets/EngMuhammadAtef/GLM-5.1-Reasoning-1M-Cleaned.carnice-glm5-hermes-traces
Carnice GLM-5 Hermes Traces
This dataset is a merged release bundle of GLM-5 traces collected through the Hermes Agent harness.
It was generated by running the carnice_trace_prompt_bank_v4 prompt bank through Hermes Agent with:
z-ai/glm-5 via OpenRouter
local/file/terminal/code-execution tools for local tasks
Hermes browser tools plus Tavily-backed web_search / web_extract for web tasks
isolated disposable workspaces per prompt
This release is prepared for Hugging Face upload and… See the full description on the dataset page: https://huggingface.co/datasets/kai-os/carnice-glm5-hermes-traces.Gemini-3.1-Pro-GLM5-CharactersPrompts generated by Gemini 3.1 Pro.
Responses generated by Gemini 3.1 Pro.
Reasoning traces:
Step 1: Generated by GLM 5 which was provided the original system prompt / knowledge
Step 2: Edited by Gemini to fix any contradictions with the existing response
Step 3: Edited again to remove / reduce drafting and remove reasoning related to safety / refusals
System prompt was generated based on the original and whatever constraints / rules etc. had been mentioned in the reasoning trace.
A small… See the full description on the dataset page: https://huggingface.co/datasets/zerofata/Gemini-3.1-Pro-GLM5-Characters.glm53-fidelity-exl3-tr3-3.42bpw-v1
fidelity--glm53.malaiwah.quant.exl3-tr3-342
A quant fidelity dataset in hidden form, produced by engines/tools/hf_capture.py from davidsyoung/GLM-5.3-EXL3-TR3-3.42bpw.
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/glm53-fidelity-exl3-tr3-3.42bpw-v1.glm-5.2-kernelgym-rollouts
GLM-5.2 KernelGym Rollouts
This dataset contains 3,200 feedback-driven GPU-kernel optimization trajectories
generated by zai-org/GLM-5.2-FP8: 100 validation tasks, two backends (inline
CUDA and Triton), and 16 rollouts per task.
Each trajectory retains the prompt/feedback message history, model responses and
reasoning, extracted kernel code, KernelGym compilation and correctness results,
profiling metadata, token usage, and stopping decision. Every published record
ended with… See the full description on the dataset page: https://huggingface.co/datasets/marin-community/glm-5.2-kernelgym-rollouts.glm53-fidelity-exl3-tr3-3.0bpw-v1
fidelity--glm53.malaiwah.quant.exl3-tr3-30
A quant fidelity dataset in hidden form, produced by engines/tools/hf_capture.py from davidsyoung/GLM-5.3-EXL3-TR3-3.0bpw.
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… See the full description on the dataset page: https://huggingface.co/datasets/malaiwah/glm53-fidelity-exl3-tr3-3.0bpw-v1.glm53-fidelity-exl3-tr3-3.25bpw-v1
fidelity--glm53.malaiwah.quant.exl3-tr3-325
A quant fidelity dataset in hidden form, produced by engines/tools/hf_capture.py from davidsyoung/GLM-5.3-EXL3-TR3-3.25bpw.
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/glm53-fidelity-exl3-tr3-3.25bpw-v1.glm52-fidelity-root-v1
fidelity--glm52.malaiwah.root.bf16
A root fidelity dataset in hidden form, produced by engines/tools/hf_capture.py from zai-org/GLM-5.2.
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 as… See the full description on the dataset page: https://huggingface.co/datasets/malaiwah/glm52-fidelity-root-v1.glm-5.1-reasoning-1m-cleaned
GLM-5.1-Reasoning-1M-Cleaned
GLM-5.1-Reasoning-1M-Cleaned is a cleaned and reformatted derivative of Kassadin88/GLM-5.1-1000000x. It preserves the original four-subset layout (main, PHD-Science, Multilingual-STEM, Math) while converting every example into a unified SFT-ready schema with explicit conversations, input, output, domain, and meta fields.
This release was prepared from the original dataset published by Kassadin88.
Summary
Teacher model in the data:… See the full description on the dataset page: https://huggingface.co/datasets/ansulev/glm-5.1-reasoning-1m-cleaned.glm5.2-general-distill
Teacher-generated instruction/response pairs used to distill small, local student models
(the ADI / Advanced Data Intelligence series) from the frontier teacher glm-5.2.
How it was built
Teacher: glm-5.2 (served via Ollama Cloud as glm-5.2:cloud), queried with
thinking/reasoning disabled so every record is a single clean final answer.
Seed prompts: databricks/databricks-dolly-15k,
filtered to remove items that require an attached context passage — the closed_qa… See the full description on the dataset page: https://huggingface.co/datasets/AdvancedDataIntelligence/glm5.2-general-distill.GLM-5.2-FP8-nemotron-codealpaca-thinking
GLM-5.2-FP8 Nemotron-CodeAlpaca Thinking Dataset
820,790 single-turn conversations generated by zai-org/GLM-5.2-FP8
with thinking enabled.
Prompt source
Rows (public)
Nemotron-Post-Training-Dataset-v2
800,944
CodeAlpaca-20k (corrected prompts, instruction + "\n\n" + input)
19,846
Total
820,790
Generation: temperature=1.0, top_p=0.95, max_tokens=24576, thinking
enabled. The CodeAlpaca prompts here include the input field.
Relationship to… See the full description on the dataset page: https://huggingface.co/datasets/JessieWei/GLM-5.2-FP8-nemotron-codealpaca-thinking.