Language Models Can Learn from Verbal Feedback Without Scalar Rewards episode artwork

EPISODE · Sep 30, 2025 · 23 MIN

Language Models Can Learn from Verbal Feedback Without Scalar Rewards

from Daily Paper Cast · host Jingwen Liang, Gengyu Wang

🤗 Upvotes: 48 | cs.CL, cs.AI, cs.LG Authors: Renjie Luo, Zichen Liu, Xiangyan Liu, Chao Du, Min Lin, Wenhu Chen, Wei Lu, Tianyu Pang Title: Language Models Can Learn from Verbal Feedback Without Scalar Rewards Arxiv: http://arxiv.org/abs/2509.22638v1 Abstract: LLMs are often trained with RL from human or AI feedback, yet such methods typically compress nuanced feedback into scalar rewards, discarding much of their richness and inducing scale imbalance. We propose treating verbal feedback as a conditioning signal. Inspired by language priors in text-to-image generation, which enable novel outputs from unseen prompts, we introduce the feedback-conditional policy (FCP). FCP learns directly from response-feedback pairs, approximating the feedback-conditional posterior through maximum likelihood training on offline data. We further develop an online bootstrapping stage where the policy generates under positive conditions and receives fresh feedback to refine itself. This reframes feedback-driven learning as conditional generation rather than reward optimization, offering a more expressive way for LLMs to directly learn from verbal feedback. Our code is available at https://github.com/sail-sg/feedback-conditional-policy.

Episode metadata supplied by the publisher feed · Published Sep 30, 2025

🤗 Upvotes: 48 | cs.CL, cs.AI, cs.LG Authors: Renjie Luo, Zichen Liu, Xiangyan Liu, Chao Du, Min Lin, Wenhu Chen, Wei Lu, Tianyu Pang Title: Language Models Can Learn from Verbal Feedback Without Scalar Rewards Arxiv: http://arxiv.org/abs/2509.22638v1 Abstract: LLMs are often trained with RL from human or AI feedback, yet such methods typically compress nuanced feedback into scalar rewards, discarding much of their richness and inducing scale imbalance. We propose treating verbal feedback as a conditioning signal. Inspired by language priors in text-to-image generation, which enable novel outputs from unseen prompts, we introduce the feedback-conditional policy (FCP). FCP learns directly from response-feedback pairs, approximating the feedback-conditional posterior through maximum likelihood training on offline data. We further develop an online bootstrapping stage where the policy generates under positive conditions and receives fresh feedback to refine itself. This reframes feedback-driven learning as conditional generation rather than reward optimization, offering a more expressive way for LLMs to directly learn from verbal feedback. Our code is available at https://github.com/sail-sg/feedback-conditional-policy.

PodParley-generated summary based on available episode metadata and transcript content.

NOW PLAYING

Language Models Can Learn from Verbal Feedback Without Scalar Rewards

0:00 23:30

No transcript for this episode yet

We transcribe on demand. Request one and we'll notify you when it's ready — usually under 10 minutes.

No similar episodes found.

Frequently Asked Questions

How long is this episode of Daily Paper Cast?

This episode is 23 minutes long.

When was this Daily Paper Cast episode published?

This episode was published on September 30, 2025.

What is this episode about?

🤗 Upvotes: 48 | cs.CL, cs.AI, cs.LG Authors: Renjie Luo, Zichen Liu, Xiangyan Liu, Chao Du, Min Lin, Wenhu Chen, Wei Lu, Tianyu Pang Title: Language Models Can Learn from Verbal...

Can I download this Daily Paper Cast episode?

Yes, you can download this episode by clicking the download button on the episode player, or subscribe to the podcast in your preferred podcast app for automatic downloads.
URL copied to clipboard!