THU-KEG/LongTraceRL
LongTraceRL Dataset Dataset Description LongTraceRL is a long-context multi-hop QA dataset designed for reinforcement learning training. Each sample contains a ~128K-token context with trajectory-based tiered distractors, a multi-hop question, a ground-truth answer, and gold rubric entities for process-level reward supervision. Dataset Details Samples: 2,815 training examples Context Length: ~128K tokens per sample Language: English Source:… See the full description on the dataset page: https://huggingface.co/datasets/THU-KEG/LongTraceRL.
| Author / organization | THU-KEG |
|---|---|
| License (as declared) | apache-2.0 |
| Downloads | 143 (as reported by Hugging Face) |
| Likes | 2 |
| Created | 27 May 2026 |
| Last updated at source | 1 June 2026 |
Popularity, last 2 days
143 now (-1 since 2026-10-16). Counts as reported by the source, snapshotted daily by Silytics.
Connected in the knowledge graph
- mradermacher/LongTraceRL-4B-GGUFModel · used to train
- mradermacher/LongTraceRL-30B-i1-GGUFModel · used to train
- mradermacher/LongTraceRL-30B-GGUFModel · used to train
- mradermacher/LongTraceRL-4B-i1-GGUFModel · used to train
Related papers
More from THU-KEG
- THU-KEG/RM-BenchDataset
- THU-KEG/WildFBDataset
- THU-KEG/AgentIFDataset
- THU-KEG/MM_MathDataset
- THU-KEG/VerInstructDataset
- THU-KEG/CaRR-DeepDiveDataset
Silytics shows public metadata from Hugging Face and links to the original. Information may be incomplete or out of date. Last refreshed 17 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.