datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
Claude-opus-5-xhigh-workload-agent-preview
Overview
Vietnamese multi-turn tool-use conversations with a <think> block on every assistant turn.
Notes: this only the preview version not fully dataset
examples
368
assistant turns
842 — 100% carry <think>
reasoning generated by
claude-opus-5
format
OpenAI-chat JSONL
Configs
from datasets import load_dataset
ds = load_dataset("beyoru/misa-agentwork-reasoning") # with <think>
ds =… See the full description on the dataset page: https://huggingface.co/datasets/beyoru/Claude-opus-5-xhigh-workload-agent-preview.swarm-llm-workload
SWARM-LLM Workload
This dataset contains the controlled question-answering workloads used by
SWARM-LLM, a prototype routing and collaboration layer for edge-hosted small
language models with optional cloud foundation model escalation.
The default train split contains the 50-query study workload reported in the
paper:
20 easy factual questions
20 hard reasoning or multi-step knowledge questions
10 safety-oriented prompts
A 5-query sanity workload used for system bring-up is… See the full description on the dataset page: https://huggingface.co/datasets/mdahshan/swarm-llm-workload.