PaulR11/training-runs-pairwise
GRPO Training Runs — Pairwise (Single-Target, With and Without FP) Part of the data release for "Training Alignment Auditors via Reinforcement Learning" (ICLR 2026). Four pairwise-reward runs. All use iterative pairwise comparison against a cached baseline transcript that is periodically refreshed from the current checkpoint. apw_grpo_anon_pw_v2_1fp/ — 1/8 FP calibration, target: Llama 3.3 70B ("Pairwise 1/8 FP") ap4_grpo_anon_pw_v2_4fp/ — 4/8 FP calibration, target: Llama 3.3… See the full description on the dataset page: https://huggingface.co/datasets/PaulR11/training-runs-pairwise.
| Author / organization | PaulR11 |
|---|---|
| License (as declared) | other |
| Downloads | 309 (as reported by Hugging Face) |
| Likes | 0 |
| Created | 24 April 2026 |
| Last updated at source | 24 April 2026 |
Popularity, last 2 days
309 now. Counts as reported by the source, snapshotted daily by Silytics.
Recent changes
- 16 Oct 2026 Description: GRPO Training Runs — Pairwise (Single-Target, With and Without FP) Part of the data release for "Training Alignment Auditors via Reinforcement Learning" (ICLR 2026). Four pairwise-reward runs. All use iterative pairwise comparison against a → GRPO Training Runs — Pairwise (Single-Target, With and Without FP) Part of the data release for "Training Alignment Auditors via Reinforcement Learning" (ICLR 2026). Four pairwise-reward runs. All use iterative pairwise comparison against a
Changes detected by Silytics between refreshes of the source.
More from PaulR11
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 Information may be incomplete or out of date.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.