arXiv 2609.38847 · cs.LG · 2026-09-30
Scoring Higher, Answering Worse: Mitigating Reward Hacking in Rubric-Based RL via Protocol-Level Rubrics
Maoqi Liu, Junwei He, Bowen Zhang, Feiran Li, Wentao Ma, Rongyi Lin, Shuhan Zhong, Quan Fang
ByteDance
In latest · #160 Added on 2026-10-21
One-line takeaway
Shows weighted-sum rubric rewards can be hacked (e.g. in clinical consultation) and proposes protocol-level rubrics to fix it.
指出基于评分细则的 RL 用加权求和会被“刷分”（以临床问诊为例），并提出协议级细则加以缓解。
Read on arXiv PDF Hugging Face discussion Code
Topics & disciplines
More on Reinforcement learning
- Waypoint-OPD: Transferring RL-Induced Policy Shifts via Gated On-Policy Distillation ▲218
- Self-Retrospection Distillation: Turning Post-hoc Experiences into Prior Foresight ▲189
- LoGRA: Scaling LLM Reinforcement Learning with Low-Rank Gradient Sketches ▲152
- Sharpening Tax in Post-Training ▲111
- RULER: Instance-aware Rubric Rewards for SVG Generation ▲104
About this data
Metadata from the arXiv API (verified); upvotes from Hugging Face Daily Papers; citations from Semantic Scholar. Last refreshed 2026-10-22.
The takeaway is written by AIMap from the abstract; the abstract itself is not reproduced — read the original.