作为一个周六的有趣“vibe coding”项目,也为了延续我之前那条推文,我临时做了一个 llm-council(LLM 议会)网页应用。 它的界面看起来和 ChatGPT 几乎一模一样,但每次用户提问都会执行以下流程: 1)使用 OpenRouter 将问题分发给议会中的多个模型,目前包括: "openai/gpt-5.1" "google/gemini-3-pro-preview" "anthropic/claude-sonnet-4.5" "x-ai/grok-4" 2)然后,这些模型都会看到彼此“匿名处理后的回答”,并对其他模型的回答进行审阅和排名。 3)最后,一个“主席模型(Chairman LLM)”会接收所有这些信息作为上下文,生成最终回复。 在同一问题下,同时看到多个模型的输出非常有趣,而更有意思的是阅读它们彼此之间如何评价和排名对方的回答。 令人意外的是,模型们经常非常坦率地认为“其他 LLM 的回答比自己的更好”,这让这种方式成为一种值得研究的模型评估策略。例如,今天我和“LLM 议会”一起阅读书籍章节时,各模型一致称赞 GPT-5.1 是最优秀、最有洞见的模型,而几乎一致认为 Claude 是表现最差的,其他模型则介于两者之间。 不过,我不完全确定这些排名是否真的符合我个人的主观感受。比如,就主观体验而言,我觉得 GPT-5.1 的回答略显冗长发散,而 Gemini 3 更加紧凑、加工度更高;Claude 在这个场景下又显得过于简短。 话虽如此,LLM 议会的数据流设计本身可能有非常大的探索空间。 LLM 模型组合(LLM ensembles)的构建方式仍然是一个尚未充分研究的领域。 As a fun Saturday vibe code project and following up on this tweet earlier, I hacked up an **llm-council** web app. It looks exactly like ChatGPT except each user query is 1) dispatched to multiple models on your council using OpenRouter, e.g. currently: "openai/gpt-5.1", "google/gemini-3-pro-preview", "anthropic/claude-sonnet-4.5", "x-ai/grok-4", Then 2) all models get to see each other's (anonymized) responses and they review and rank them, and then 3) a "Chairman LLM" gets all of that as context and produces the final response. It's interesting to see the results from multiple models side by side on the same query, and even more amusingly, to read through their evaluation and ranking of each other's responses. Quite often, the models are surprisingly willing to select another LLM's response as superior to their own, making this an interesting model evaluation strategy more generally. For example, reading book chapters together with my LLM Council today, the models consistently praise GPT 5.1 as the best and most insightful model, and consistently select Claude as the worst model, with the other models floating in between. But I'm not 100% convinced this aligns with my own qualitative assessment. For example, qualitatively I find GPT 5.1 a little too wordy and sprawled and Gemini 3 a bit more condensed and processed. Claude is too terse in this domain. That said, there's probably a whole design space of the data flow of your LLM council. The construction of LLM ensembles seems under-explored.

Andrej KarpathyFounder, Eureka Labs AI

观点原文

作为一个周六的有趣“vibe coding”项目,也为了延续我之前那条推文,我临时做了一个 llm-council(LLM 议会)网页应用。

它的界面看起来和 ChatGPT 几乎一模一样,但每次用户提问都会执行以下流程:

1)使用 OpenRouter 将问题分发给议会中的多个模型,目前包括:

"openai/gpt-5.1"

"google/gemini-3-pro-preview"

"anthropic/claude-sonnet-4.5"

"x-ai/grok-4"

2)然后,这些模型都会看到彼此“匿名处理后的回答”,并对其他模型的回答进行审阅和排名。

3)最后,一个“主席模型(Chairman LLM)”会接收所有这些信息作为上下文,生成最终回复。

在同一问题下,同时看到多个模型的输出非常有趣,而更有意思的是阅读它们彼此之间如何评价和排名对方的回答。

令人意外的是,模型们经常非常坦率地认为“其他 LLM 的回答比自己的更好”,这让这种方式成为一种值得研究的模型评估策略。例如,今天我和“LLM 议会”一起阅读书籍章节时,各模型一致称赞 GPT-5.1 是最优秀、最有洞见的模型,而几乎一致认为 Claude 是表现最差的,其他模型则介于两者之间。

不过,我不完全确定这些排名是否真的符合我个人的主观感受。比如,就主观体验而言,我觉得 GPT-5.1 的回答略显冗长发散,而 Gemini 3 更加紧凑、加工度更高;Claude 在这个场景下又显得过于简短。

话虽如此,LLM 议会的数据流设计本身可能有非常大的探索空间。

LLM 模型组合(LLM ensembles)的构建方式仍然是一个尚未充分研究的领域。

As a fun Saturday vibe code project and following up on this tweet earlier, I hacked up an llm-council web app. It looks exactly like ChatGPT except each user query is 1) dispatched to multiple models on your council using OpenRouter, e.g. currently:

"openai/gpt-5.1",

"google/gemini-3-pro-preview",

"anthropic/claude-sonnet-4.5",

"x-ai/grok-4",

Then 2) all models get to see each other's (anonymized) responses and they review and rank them, and then 3) a "Chairman LLM" gets all of that as context and produces the final response.

It's interesting to see the results from multiple models side by side on the same query, and even more amusingly, to read through their evaluation and ranking of each other's responses.

Quite often, the models are surprisingly willing to select another LLM's response as superior to their own, making this an interesting model evaluation strategy more generally. For example, reading book chapters together with my LLM Council today, the models consistently praise GPT 5.1 as the best and most insightful model, and consistently select Claude as the worst model, with the other models floating in between. But I'm not 100% convinced this aligns with my own qualitative assessment. For example, qualitatively I find GPT 5.1 a little too wordy and sprawled and Gemini 3 a bit more condensed and processed. Claude is too terse in this domain.

That said, there's probably a whole design space of the data flow of your LLM council. The construction of LLM ensembles seems under-explored.

AI

相关观点