Verbal Agent Reinforcement

AdviSD adds a selective feedback mechanism to the earlier verbal reinforcement line. A small advisor proposes natural-language corrections to a frozen executor, but the method avoids treating every plausible reflection as useful training signal: advice that does not improve execution can constrain later learning. The accumulated line now contains several independent research approaches across multiple days, including user feedback and learning from failed attempts. It reaches the stored publication threshold, but the commercial interpretation remains emerging: useful feedback needs selection and verification, not merely collection. Production improvement and resistance to harmful feedback remain open questions.

The MURPHY work trains models from the complete process of attempting, failing, receiving feedback and correcting themselves rather than only from final successful outputs. It is a direct continuation of the feedback-driven agent-learning line, but remains research-led. Production evidence showing durable improvement without accumulating bad feedback would be required for publication.

ScienceBuddy explicitly transforms researcher requests and feedback into tasks and evaluation rubrics for continual learning. This is a concrete implementation of the existing verbal-feedback hypothesis, but it remains research-led. Independent production learning loops with measured improvement and safeguards against bad feedback would confirm the market consequence.

A second research result argues that naturally occurring user feedback contains improvement information that models cannot reliably infer from the interaction alone. This strengthens the existing hypothesis that explanations, corrections and complaints can become a reinforcement channel, but both observations remain research-led. An open implementation or a measured production feedback loop would be needed for publication.