Robust Reward Modeling via Causal Rubrics

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Srivastava, Pragya, Singh, Harman, Madhavan, Rahul, Patil, Gandharv, Addepalli, Sravanti, Suggala, Arun, Aravamudhan, Rengarajan, Sharma, Soumya, Laha, Anirban, Raghuveer, Aravindan, Shanmugam, Karthikeyan, Precup, Doina
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866909653987753984
author Srivastava, Pragya
Singh, Harman
Madhavan, Rahul
Patil, Gandharv
Addepalli, Sravanti
Suggala, Arun
Aravamudhan, Rengarajan
Sharma, Soumya
Laha, Anirban
Raghuveer, Aravindan
Shanmugam, Karthikeyan
Precup, Doina
author_facet Srivastava, Pragya
Singh, Harman
Madhavan, Rahul
Patil, Gandharv
Addepalli, Sravanti
Suggala, Arun
Aravamudhan, Rengarajan
Sharma, Soumya
Laha, Anirban
Raghuveer, Aravindan
Shanmugam, Karthikeyan
Precup, Doina
contents Reward models (RMs) are fundamental to aligning Large Language Models (LLMs) via human feedback, yet they often suffer from reward hacking. They tend to latch on to superficial or spurious attributes, such as response length or formatting, mistaking these cues learned from correlations in training data for the true causal drivers of quality (e.g., factuality, relevance). This occurs because standard training objectives struggle to disentangle these factors, leading to brittle RMs and misaligned policies. We introduce Crome (Causally Robust Reward Modeling), a novel framework grounded in an explicit causal model designed to mitigate reward hacking. Crome employs the following synthetic targeted augmentations during training: (1) Causal Augmentations, which are pairs that differ along specific causal attributes, to enforce sensitivity along each causal attribute individually, and (2) Neutral Augmentations, which are tie-label pairs varying primarily in spurious attributes, to enforce invariance along spurious attributes. Notably, our augmentations are produced without any knowledge of spurious factors, via answer interventions only along causal rubrics, that are identified by querying an oracle LLM. Empirically, Crome significantly outperforms standard baselines on RewardBench, improving average accuracy by up to 5.4% and achieving gains of up to 13.2% and 7.2% in specific categories. The robustness of Crome is further testified by the consistent gains obtained in a Best-of-N inference setting across increasing N, across various benchmarks, including the popular RewardBench (covering chat, chat-hard, safety, and reasoning tasks), the safety-focused WildGuardTest, and the reasoning-specific GSM8k.
format Preprint
id arxiv_https___arxiv_org_abs_2506_16507
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Robust Reward Modeling via Causal Rubrics
Srivastava, Pragya
Singh, Harman
Madhavan, Rahul
Patil, Gandharv
Addepalli, Sravanti
Suggala, Arun
Aravamudhan, Rengarajan
Sharma, Soumya
Laha, Anirban
Raghuveer, Aravindan
Shanmugam, Karthikeyan
Precup, Doina
Machine Learning
Reward models (RMs) are fundamental to aligning Large Language Models (LLMs) via human feedback, yet they often suffer from reward hacking. They tend to latch on to superficial or spurious attributes, such as response length or formatting, mistaking these cues learned from correlations in training data for the true causal drivers of quality (e.g., factuality, relevance). This occurs because standard training objectives struggle to disentangle these factors, leading to brittle RMs and misaligned policies. We introduce Crome (Causally Robust Reward Modeling), a novel framework grounded in an explicit causal model designed to mitigate reward hacking. Crome employs the following synthetic targeted augmentations during training: (1) Causal Augmentations, which are pairs that differ along specific causal attributes, to enforce sensitivity along each causal attribute individually, and (2) Neutral Augmentations, which are tie-label pairs varying primarily in spurious attributes, to enforce invariance along spurious attributes. Notably, our augmentations are produced without any knowledge of spurious factors, via answer interventions only along causal rubrics, that are identified by querying an oracle LLM. Empirically, Crome significantly outperforms standard baselines on RewardBench, improving average accuracy by up to 5.4% and achieving gains of up to 13.2% and 7.2% in specific categories. The robustness of Crome is further testified by the consistent gains obtained in a Best-of-N inference setting across increasing N, across various benchmarks, including the popular RewardBench (covering chat, chat-hard, safety, and reasoning tasks), the safety-focused WildGuardTest, and the reasoning-specific GSM8k.
title Robust Reward Modeling via Causal Rubrics
topic Machine Learning
url https://arxiv.org/abs/2506.16507