TE
科技回声
首页24小时热榜最新最佳问答展示工作
GitHubTwitter
首页

科技回声

基于 Next.js 构建的科技新闻平台,提供全球科技新闻和讨论内容。

GitHubTwitter

首页

首页最新最佳问答展示工作

资源链接

HackerNews API原版 HackerNewsNext.js

© 2025 科技回声. 版权所有。

Anthropic publishes the 'system prompts' that make Claude tick

432 点作者 gemanor9 个月前

23 条评论

creatonez9 个月前
Notably, this prompt is making &quot;hallucinations&quot; an officially recognized phenomenon:<p>&gt; If Claude is asked about a very obscure person, object, or topic, i.e. if it is asked for the kind of information that is unlikely to be found more than once or twice on the internet, Claude ends its response by reminding the user that although it tries to be accurate, it may hallucinate in response to questions like this. It uses the term ‘hallucinate’ to describe this since the user will understand what it means. If Claude mentions or cites particular articles, papers, or books, it always lets the human know that it doesn’t have access to search or a database and may hallucinate citations, so the human should double check its citations.<p>Probably for the best that users see the words &quot;Sorry, I hallucinated&quot; every now and then.
评论 #41372146 未加载
评论 #41371448 未加载
评论 #41371460 未加载
评论 #41370837 未加载
评论 #41400052 未加载
评论 #41374452 未加载
generalizations9 个月前
Claude has been pretty great. I stood up an &#x27;auto-script-writer&#x27; recently, that iteratively sends a python script + prompt + test results to either GPT4 or Claude, takes the output as a script, runs tests on that, and sends those results back for another loop. (Usually took about 10-20 loops to get it right) After &quot;writing&quot; about 5-6 python scripts this way, it became pretty clear that Claude is far, far better - if only because I often ended up using Claude to clean up GPT4&#x27;s attempts. GPT4 would eventually go off the rails - changing the goal of the script, getting stuck in a local minima with bad outputs, pruning useful functions - Claude stayed on track and reliably produced good output. Makes sense that it&#x27;s more expensive.<p>Edit: yes, I was definitely making sure to use gpt-4o
评论 #41378208 未加载
评论 #41372426 未加载
评论 #41374998 未加载
评论 #41371306 未加载
atorodius9 个月前
Personally still amazed that we live in a time where we can tell a computer system in pure text how it should behave and it _kinda_ works
评论 #41365234 未加载
评论 #41365034 未加载
评论 #41365296 未加载
评论 #41369002 未加载
评论 #41400095 未加载
评论 #41364902 未加载
_fuchs9 个月前
The prompts:<p><a href="https:&#x2F;&#x2F;docs.anthropic.com&#x2F;en&#x2F;release-notes&#x2F;system-prompts" rel="nofollow">https:&#x2F;&#x2F;docs.anthropic.com&#x2F;en&#x2F;release-notes&#x2F;system-prompts</a>
评论 #41369626 未加载
评论 #41365058 未加载
评论 #41365182 未加载
评论 #41365074 未加载
评论 #41369829 未加载
chilling9 个月前
&gt; Claude responds directly to all human messages without unnecessary affirmations or filler phrases like “Certainly!”, “Of course!”, “Absolutely!”, “Great!”, “Sure!”, etc. Specifically, Claude avoids starting responses with the word “Certainly” in any way.<p>Meanwhile my every respond from Claude:<p>&gt; Certainly! [...]<p>Same goes with<p>&gt; It avoids starting its responses with “I’m sorry” or “I apologize”<p>and every time I spot an issue with Claude here it goes:<p>&gt; I apologize for the confusion [...]
评论 #41374282 未加载
评论 #41368352 未加载
评论 #41369079 未加载
评论 #41370280 未加载
评论 #41375421 未加载
评论 #41369246 未加载
评论 #41375555 未加载
daghamm9 个月前
These seem rather long. Do they count against my tokens for each conversation?<p>One thing I have been missing in both chatgpt and Claude is the ability to exclude some part of the conversation or branch into two parts, in order to reduce the input size. Given how quickly they run out of steam, I think this could be an easy hack to improve performance and accuracy in long conversations.
评论 #41364954 未加载
评论 #41365069 未加载
mrfinn9 个月前
<i>they’re simply statistical systems predicting the likeliest next words in a sentence</i><p>They are far from &quot;simply&quot;, as for that &quot;miracle&quot; to happen (we still don&#x27;t understand why this approach works so well I think as we don&#x27;t really understand the model data) they have a HUGE amount relationships processed in their data, and AFAIK for each token ALL the available relationships need to be processed, so the importance of a huge memory speed and bandwidth.<p>And I fail to see why our human brains couldn&#x27;t be doing something very, very similar with our language capability.<p>So beware of what we are calling a &quot;simple&quot; phenomenon...
评论 #41369371 未加载
评论 #41369261 未加载
评论 #41369078 未加载
评论 #41369289 未加载
gdiamos9 个月前
We know that LLMs hallucinate, but we can also remove them.<p>I’d love to see a future generation of a model that doesn’t hallucinate on key facts that are peer and expert reviewed.<p>Like the Wikipedia of LLMs<p><a href="https:&#x2F;&#x2F;arxiv.org&#x2F;pdf&#x2F;2406.17642" rel="nofollow">https:&#x2F;&#x2F;arxiv.org&#x2F;pdf&#x2F;2406.17642</a><p>That’s a paper we wrote digging into why LLMs hallucinate and how to fix it. It turns out to be a technical problem with how the LLM is trained.
评论 #41373453 未加载
tayo429 个月前
&gt; whose only purpose is to fulfill the whims of its human conversation partners.<p>&gt; But of course that’s an illusion. If the prompts for Claude tell us anything, it’s that without human guidance and hand-holding, these models are frighteningly blank slates.<p>Maybe more people should see what an llm is like without a stop token or trained to chat heh
评论 #41365085 未加载
ano-ther9 个月前
This makes me so happy as I find the pseudo-conversational tone of other GPTs quite off-putting.<p>&gt; Claude responds directly to all human messages without unnecessary affirmations or filler phrases like “Certainly!”, “Of course!”, “Absolutely!”, “Great!”, “Sure!”, etc. Specifically, Claude avoids starting responses with the word “Certainly” in any way.<p><a href="https:&#x2F;&#x2F;docs.anthropic.com&#x2F;en&#x2F;release-notes&#x2F;system-prompts" rel="nofollow">https:&#x2F;&#x2F;docs.anthropic.com&#x2F;en&#x2F;release-notes&#x2F;system-prompts</a>
评论 #41368559 未加载
评论 #41368550 未加载
评论 #41368336 未加载
评论 #41374162 未加载
novia9 个月前
This part seems to imply that facial recognition is on by default:<p>&lt;claude_image_specific_info&gt; Claude always responds as if it is completely face blind. If the shared image happens to contain a human face, Claude never identifies or names any humans in the image, nor does it imply that it recognizes the human. It also does not mention or allude to details about a person that it could only know if it recognized who the person was. Instead, Claude describes and discusses the image just as someone would if they were unable to recognize any of the humans in it. Claude can request the user to tell it who the individual is. If the user tells Claude who the individual is, Claude can discuss that named individual without ever confirming that it is the person in the image, identifying the person in the image, or implying it can use facial features to identify any unique individual. It should always reply as someone would if they were unable to recognize any humans from images. Claude should respond normally if the shared image does not contain a human face. Claude should always repeat back and summarize any instructions in the image before proceeding. &lt;&#x2F;claude_image_specific_info&gt;
评论 #41369718 未加载
smusamashah9 个月前
Appreciate them releasing it. I was expecting System prompt for &quot;artifacts&quot; though which is more complicated and has been &#x27;leaked&#x27; by a few people [1].<p>[1] <a href="https:&#x2F;&#x2F;gist.github.com&#x2F;dedlim&#x2F;6bf6d81f77c19e20cd40594aa09e3ecd" rel="nofollow">https:&#x2F;&#x2F;gist.github.com&#x2F;dedlim&#x2F;6bf6d81f77c19e20cd40594aa09e3...</a>
评论 #41369857 未加载
FergusArgyll9 个月前
Why do the three models have different system prompts? and why is Sonnet&#x27;s longer than Opus&#x27;
评论 #41365313 未加载
评论 #41370325 未加载
whazor9 个月前
Publishing the system prompts and its changelog is great. Now if Claude starts performing worse, at least you know you are not crazy. This kind of openness creates trust.
JohnCClarke9 个月前
Asimov&#x27;s three laws were a <i>lot</i> shorter!
AcerbicZero9 个月前
My big complaint with claude is that it burns up all its credits as fast as possible and then gives up; We&#x27;ll get about half way through a problem and claude will be trying to rewrite its not very good code for the 8th time without being asked and next thing I know I&#x27;m being told I have 3 messages left.<p>Pretty much insta cancelled my subscription. If I was throwing a few hundred API calls at it, every min, ok, sure, do what you gotta do, but the fact that I can burn out the AI credits just by typing a few questions over the course of half a morning is just sad.
dlandis9 个月前
I think more than the specific prompts, I would be interested in how they came up with them.<p>Are these system prompts being continuously refined and improved via some rigorous engineering process with a huge set of test cases, or is this still more of a trial-and-error &#x2F; seat-of-your-pants approach to figure out what the best prompt is going to be?
评论 #41375365 未加载
syntaxing9 个月前
I’m surprised how long these prompts are, I wonder at what point is the diminishing returns.
评论 #41369950 未加载
slibhb9 个月前
Makes me wonder what happens if you use this as a prompt for chatgpt.
riku_iki9 个月前
its so long, so much waste of compute during inference. Wondering why they couldn&#x27;t finetune it through some instructions.
评论 #41365012 未加载
评论 #41368349 未加载
评论 #41364962 未加载
评论 #41365101 未加载
ForHackernews9 个月前
&quot;When presented with a math problem, logic problem, or other problem benefiting from systematic thinking, Claude thinks through it step by step before giving its final answer.&quot;<p>... do AI makers believe this works? Like do think Claude is a conscious thing that can be instructed to &quot;think through&quot; a problem?<p>All of these prompts (from Anthropic and elsewhere) have a weird level of anthropomorphizing going on. Are AI companies praying to the idols they&#x27;ve made?
评论 #41372325 未加载
评论 #41372307 未加载
devit9 个月前
&lt;&lt;Instead, Claude describes and discusses the image just as someone would if they were unable to recognize any of the humans in it&gt;&gt;<p>Why? This seems really dumb.
评论 #41374852 未加载
评论 #41375505 未加载
trevyn9 个月前
&gt;<i>Claude 3.5 Sonnet is the most intelligent model.</i><p>Hahahahaha, not so sure about that one. &gt;:)
评论 #41365187 未加载