Reference library
Training language models to follow instructions with human feedback
Paper summary
InstructGPT combines demonstration-based fine-tuning, learned human-preference rewards, and reinforcement learning. Human evaluations show improved alignment with user intent compared with larger untreated language models in the tested setting.
Relevance to the project
A clear behavioral-origin example: capability and evaluations reflect explicit human demonstrations and preferences. Useful in assessing ordinary LLMs under the properties catalog.
Human preference alignment is distinct from internally arising recurrence-supporting behavior.
Source record
- Reference ID
- arxiv-2203-02155
- Archived version
- 2203.02155v1
- Summary provenance
- reference-library/papers/arxiv-2203-02155/metadata.json · summary
- Review status
- initial orientation summary; not a critical full-paper review
- Reading scope
- Selected abstract/opening and available closing discussion/conclusion passages in the preserved text extract; broader project notes provide context. No reproduction of experiments, exhaustive methods review, or equation/figure audit.