科技回声

2 条评论

azaras大约 2 个月前

It is interesting for an old and classic software developer to see how the LLMs are taught/programmed. They use markdown files:<p>- <a href="https://github.com/SimonAytes/SoT/blob/main/sketch_of_thought/config/prompts/EN/EN_CoT_SystemPrompt.md" rel="nofollow">https://github.com/SimonAytes/SoT/blob/main/sketch_of_though...</a><p>- <a href="https://github.com/SimonAytes/SoT/blob/main/sketch_of_thought/config/prompts/EN/EN_ConceptualChaining_SystemPrompt.md" rel="nofollow">https://github.com/SimonAytes/SoT/blob/main/sketch_of_though...</a><p>- <a href="https://github.com/SimonAytes/SoT/blob/main/sketch_of_thought/config/prompts/EN/EN_ExpertLexicons_SystemPrompt.md" rel="nofollow">https://github.com/SimonAytes/SoT/blob/main/sketch_of_though...</a>

评论 #43410779 未加载

Bolwin大约 2 个月前

This seems near identical to chain of draft, albeit more thorough and they actually did training instead of just promoting <a href="https://arxiv.org/abs/2502.18600" rel="nofollow">https://arxiv.org/abs/2502.18600</a>

评论 #43420295 未加载

Sketch-of-Thought: Efficient LLM Reasoning

2 条评论

Sketch-of-Thought: Efficient LLM Reasoning

2 条评论