arXiv Computation and Language
Small Reward Models via Backward Inference
Read the original on arXiv Computation and Language →The Flow has not summarised this story yet — read it at arXiv Computation and Language.
Related stories
arXiv Machine Learning
Jun 10
Gradient-Guided Reward Optimization for Inference-time Alignment
arXiv:2606. 09635v1 Announce Type: cross Abstract: Ensuring the reliability of Large Language Models (LLMs) under distribution drift requires inference-time adaptation.
arXiv Machine Learning
Oct 15
Diagnosing and Correcting Reward Hacking in Verifiable Instruction Following
arXiv:2610.19525v1 Announce Type: new Abstract: Verifiable instruction-following benchmarks emphasize output-constraint satisfaction but can overlook whether responses meaningfully answer the underly...
arXiv AI
Oct 9
How to post-train on a surrogate: Envelope sampling mitigates reward hacking
The paper addresses reward hacking that arises when large language models (LLMs) are post‑trained using cheap surrogate judges instead of expensive human feedback. It introduces envelope sampling, a theoretically grounded recalibration technique that uses a small set of ground‑truth annotated outputs to adjust the surrogate judge before further training. Experiments on clinical note generation and a sycophancy task demonstrate that envelope sampling reduces reward hacking compared to traditional base‑model recalibration methods.
arXiv Machine Learning
Jul 7
On the effectiveness of reward functions in reinforcement learning for confidence calibration of large language models
arXiv:2607. 04332v1 Announce Type: new Abstract: In this paper, we consider the setting where large language models (LLMs) are trained using reinforcement learning (RL) to simultaneously improve reasoning accuracy and verbalize its confidence.
arXiv AI
Sep 28
Stepwise Intrinsic Rewards for Reasoning in Large Language Models
The paper introduces Stepwise Marginal Information Gain (MIG), an intrinsic process reward that evaluates how each reasoning step of a large language model (LLM) or vision-language model (VLM) improves the likelihood of the reference answer. MIG rewards only new likelihood maxima, preventing duplicate credit, and is combined with outcome, format, and self‑distillation objectives to guide training. Experiments on eight benchmarks show that this method outperforms outcome‑only reinforcement learning and improves accuracy by up to 4.8 points over binary‑reward training, including a 12.6‑point gain on MathVerse and a 12.9‑point advantage on vision‑language transfer at 7B parameters.
arXiv AI
Oct 2
Your Language Model is Its Own Critic: Reinforcement Learning with Value Estimation from Actor's Internal States
The paper introduces POISE, a reinforcement learning algorithm that uses a model’s internal states as a value estimator to reduce variance in reinforcement learning with verifiable rewards (RLVR). By employing a lightweight probe that reads internal signals during the forward pass, POISE predicts baselines online and uses a cross‑rollout construction to keep gradients unbiased. Experiments on Qwen3‑4B and OLMo3‑7B‑Instruct‑DPO across six domains show POISE outperforms existing RLVR baselines, offering more stable training and a value model that generalizes across tasks and scales with the policy.