TE
TechEcho
Home24h TopNewestBestAskShowJobs
GitHubTwitter
Home

TechEcho

A tech news platform built with Next.js, providing global tech news and discussions.

GitHubTwitter

Home

HomeNewestBestAskShowJobs

Resources

HackerNews APIOriginal HackerNewsNext.js

© 2025 TechEcho. All rights reserved.

LLMs Are Interpretable

17 pointsby tkelloggover 1 year ago

5 comments

imranqover 1 year ago
I think interpretable is a overloaded term. If you use RAG, sure you can look at the retrieved text and understand what is being used, but at that point you&#x27;re just interpreting the retriever.<p>LLMs need to be interpreted so that they can be edited have their biases understood in a systematic way. However just as people aren&#x27;t &quot;interpretable&quot; these algorithms are not going to be able to display their inner workings with 100% confidence. It&#x27;s going to remain probabilistic, which might be fine for the majority of use cases. I think we&#x27;re coming from an age where everything was 100% interpretable because we knew what was going on inside the machine (e.g. in a knowledge graph).<p>There needs to be some definition of what we want to achieve with interpretability for us to understand what standards we need to keep.
评论 #37779394 未加载
lukevover 1 year ago
I like this point of view, and I&#x27;ll even go a step further...<p>If we&#x27;re going to use LLMs as the basis for anything resembling general intelligence, it won&#x27;t be through one-shot invocation of the model. It&#x27;ll be through some kind of chain&#x2F;tree&#x2F;graph of thought where the model invokes itself recursively.<p>In this scenario, we have an exact transcript of the model&#x27;s thought process, exactly as it occurred, <i>written for us in english.</i> The model can&#x27;t even have a private &quot;thought&quot;, everything needs to be in the visible context.<p>You can&#x27;t get more interpretable than that.
zbyover 1 year ago
A great article - just a technical nitpick for the author: &quot;in-context learning&quot; is not what you do with RAG. &quot;In context learning&quot; is a really confusing name for reasoning by analogy. In RAG you provide source information in the prompt - in ICL you provide examples of how the task should be accomplished. :<p>&quot;&quot;&quot;<p>In-context learning in language models, also known as few-shot learning or few-shot prompting, is a technique where the model is presented with prompts and responses as a context prior to performing a task. For example, to train a language model to generate imaginative and witty jokes.<p>We can leverage in-context learning by exposing the model to a dataset of joke prompts and corresponding punchlines:<p><pre><code> Prompt 1: “Why don’t scientists trust atoms?” Response: “Because they make up everything! Prompt 2: “What do you call a bear with no teeth?” Response: “A gummy bear!” Prompt 3: “Why did the scarecrow win an award?” Response: “Because he was outstanding in his field!” </code></pre> By training in different types of jokes, the model develops an understanding of how humor works and becomes capable of creating its own clever and amusing punchlines.<p>&quot;&quot;&quot;<p>from <a href="https:&#x2F;&#x2F;www.techopedia.com&#x2F;from-language-models-to-problem-solvers-the-rise-of-in-context-learning-in-ais-problem-solving-journey" rel="nofollow noreferrer">https:&#x2F;&#x2F;www.techopedia.com&#x2F;from-language-models-to-problem-s...</a>
评论 #37783625 未加载
sirwhinesalotover 1 year ago
If you consider your schizophrenic uncle to be &quot;interpretable&quot;, sure.
评论 #37780115 未加载
gorenbover 1 year ago
must read for anyone interested in llms