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。

Hermes 3 – Nous Research

29 点作者 jasim6 个月前

2 条评论

vintagedave6 个月前
It&#x27;s a fine-tuned model over Llama 3.1, which is the kind of thing I -- a non-expert -- would want to do and have though of doing if I trained a LLM for a specific programming language for private (non-cloud-hosted) use. So this is both interesting and yet I lack to knowledge to really understand its impact in the LLM world.<p>&gt; the model displays significant improvements in judgment and reward modeling.<p>This seems significant?<p>Technical report: <a href="https:&#x2F;&#x2F;nousresearch.com&#x2F;wp-content&#x2F;uploads&#x2F;2024&#x2F;08&#x2F;Hermes-3-Technical-Report.pdf" rel="nofollow">https:&#x2F;&#x2F;nousresearch.com&#x2F;wp-content&#x2F;uploads&#x2F;2024&#x2F;08&#x2F;Hermes-3...</a>
评论 #42040123 未加载
simonw6 个月前
The documentation is missing some details: <a href="https:&#x2F;&#x2F;docs.lambdalabs.com&#x2F;public-cloud&#x2F;lambda-chat-api&#x2F;" rel="nofollow">https:&#x2F;&#x2F;docs.lambdalabs.com&#x2F;public-cloud&#x2F;lambda-chat-api&#x2F;</a><p>When I run a prompt through it I get this back:<p><pre><code> { &quot;id&quot;: &quot;chat-dea5c8eddcfa4ad08d488f2501f1b3b4&quot;, &quot;object&quot;: &quot;chat.completion&quot;, &quot;created&quot;: 1730717593, &quot;model&quot;: &quot;hermes3-405b&quot;, &quot;choices&quot;: [ { &quot;index&quot;: 0, &quot;message&quot;: { &quot;role&quot;: &quot;assistant&quot;, &quot;content&quot;: &quot;The 2020 World Series was played at Globe Life Field in Arlington, Texas. Due to the COVID-19 pandemic, the entire series was held at this neutral site to reduce travel and potential exposure to the virus.&quot; }, &quot;finish_reason&quot;: &quot;stop&quot;, &quot;content_filter_results&quot;: { &quot;hate&quot;: { &quot;filtered&quot;: false }, &quot;self_harm&quot;: { &quot;filtered&quot;: false }, &quot;sexual&quot;: { &quot;filtered&quot;: false }, &quot;violence&quot;: { &quot;filtered&quot;: false }, &quot;jailbreak&quot;: { &quot;filtered&quot;: false, &quot;detected&quot;: false }, &quot;profanity&quot;: { &quot;filtered&quot;: false, &quot;detected&quot;: false } } } ], &quot;usage&quot;: { &quot;prompt_tokens&quot;: 65, &quot;completion_tokens&quot;: 45, &quot;total_tokens&quot;: 110, &quot;prompt_tokens_details&quot;: null, &quot;completion_tokens_details&quot;: null }, &quot;system_fingerprint&quot;: &quot;&quot; } </code></pre> Those content_filter_results look interesting - especially if I can turn those options on or off (I&#x27;d like to experiment with the jailbreak one for example) - but they aren&#x27;t mentioned in the documentation at the moment.