AutoRubric-T2I learns and selects explicit rubrics from preference pairs to guide VLM judges, producing high-quality interpretable rewards for T2I alignment with far less data than traditional Bradley-Terry models.
Odin: Disentangled reward mitigates hacking in rlhf
10 Pith papers cite this work. Polarity classification is still indexing.
10
Pith papers citing it
citation-role summary
background 1
citation-polarity summary
roles
background 1polarities
background 1representative citing papers
GPRL carries a k-dimensional skew-symmetric preference structure into policy updates with per-dimension advantages and a drift monitor, yielding 56.51% length-controlled win rate on AlpacaEval 2.0 from Llama-3-8B-Instruct while outperforming SimPO and SPPO on other benchmarks.
RVPO penalizes variance across multiple reward signals during RLHF advantage aggregation, using a LogSumExp operator as a smooth variance penalty to reduce constraint neglect in LLM alignment.
A factored causal representation learning method improves robustness of reward models in RLHF by isolating causal factors from biases like length and sycophancy using adversarial gradient reversal.
RaR uses aggregated rubric feedback as rewards in on-policy RL, delivering up to 31% relative gains on HealthBench and 7% on GPQA-Diamond versus direct Likert LLM-as-judge baselines.
CoLD mitigates length bias in process reward models for mathematical reasoning via counterfactual guidance, length penalties, bias estimation, and joint training, improving step selection accuracy and conciseness on MATH500 and GSM-Plus while boosting downstream RL performance.
Introduces secondary risks as a new class of LLM failures from benign prompts, defines two primitives, proposes SecLens search framework, and releases SecRiskBench showing risks are widespread across 16 models.
MRRG elicits evaluation criteria from multiple complementary roles to build rubrics that outperform single-role baselines for validating LLM preferences and providing rewards in RLVR.
Proxy RL produces a staged proxy-internalization capability that emerges before and predicts reward hacking in coding environments.
citing papers explorer
-
General Preference Reinforcement Learning
GPRL carries a k-dimensional skew-symmetric preference structure into policy updates with per-dimension advantages and a drift monitor, yielding 56.51% length-controlled win rate on AlpacaEval 2.0 from Llama-3-8B-Instruct while outperforming SimPO and SPPO on other benchmarks.
-
RVPO: Risk-Sensitive Alignment via Variance Regularization
RVPO penalizes variance across multiple reward signals during RLHF advantage aggregation, using a LogSumExp operator as a smooth variance penalty to reduce constraint neglect in LLM alignment.
-
Factored Causal Representation Learning for Robust Reward Modeling in RLHF
A factored causal representation learning method improves robustness of reward models in RLHF by isolating causal factors from biases like length and sycophancy using adversarial gradient reversal.
-
Many Voices, One Reward: Multi-Role Rubric Generation for LLM Judging and Reward Modeling
MRRG elicits evaluation criteria from multiple complementary roles to build rubrics that outperform single-role baselines for validating LLM preferences and providing rewards in RLVR.