When a Model Knows the Requirements but Fails to Follow Them: What Research on Post-Training, Character Fidelity, and Long Context Can Tell Us About GPT's Quality

🌐 链接: https://community.openai.com/t/when-a-model-knows-the-requirements-but-fails-to-follow-them-what-research-on-post-training-character-fidelity-and-long-context-can-tell-us-about-gpts-quality/1415874
🏷️ 分组: OpenAI Developer Community
🕒 时间: 2026-11-13 19:57:38
OpenAI Developer Community
When a Model Knows the Requirements but Fails to Follow Them: What Research on Post-Training, Character Fidelity, and Long Context…
Hello everyone! I’m returning to the subject of creative writing, but this time I’d like to continue our discussion from a more research-oriented perspective. Over the past few months, we’ve already talked extensively about what happens to characters in…