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.

Kagi LLM Benchmarking Project

64 pointsby bx37610 months ago

5 comments

xianshou10 months ago
This is <i>almost</i> perfect.<p>The gold standard for LLM evaluation would have the following qualities:<p>1. Categorized (e.g. coding, reasoning, general knowledge)<p>2. Multimodal (at least text and image)<p>3. Multiple difficulties (something like &quot;GPT-4 saturates or scores &gt;90%&quot;, a la MMLU, &quot;GPT-4 scores 20-80%&quot;, and &quot;GPT-4 scores &lt; 10%&quot;)<p>4. Hidden (under 10% of the dataset publicly available, enough methodological detail to inspire confidence but not enough to design to the test set)<p>The standard model card suite with MMLU, HumanEval etc. has already been optimized to the point of diminishing value - Goodhart&#x27;s law in action. Meanwhile, arena Elo (<a href="https:&#x2F;&#x2F;huggingface.co&#x2F;spaces&#x2F;lmsys&#x2F;chatbot-arena-leaderboard" rel="nofollow">https:&#x2F;&#x2F;huggingface.co&#x2F;spaces&#x2F;lmsys&#x2F;chatbot-arena-leaderboar...</a>) is extremely useful, but also has the drawback of reflecting median-voter preferences that will not necessarily correlate with true intelligence as capabilities continue to advance, in the same sense as how the doctor with the best bedside manner is not necessarily the best doctor.<p>Until that happens, I&#x27;ll pay attention to every eval I can find, but am also stuck asking &quot;how many r&#x27;s are in strawberry?&quot; and &quot;draw a 7-sided stop sign&quot; to get a general impression of intelligence independent of gameable or overly general benchmarks.<p>But all that aside:<p><pre><code> Model | Score </code></pre> ----------------------------------------------<p><pre><code> GPT-4o | 52 Llama 3.1 405B | 50 Claude 3.5 Sonnet | 46 Mistral Large | 44 Gemini 1.5 Pro | 12 </code></pre> What an incredible contrast to MMLU, where all of these models score in the 80-90% range! For what it&#x27;s worth, these scores also fall much closer to my impressions from daily use. Gemini is awful, Sonnet and 4o are amazing, and the new Llama puts fine-tunable, open-source 4o in the hands of anyone with a mini-cluster.
评论 #41112419 未加载
评论 #41112083 未加载
评论 #41112007 未加载
评论 #41122220 未加载
tunesmith10 months ago
Having very little familiarity of models other than using ChatGPT+, I can&#x27;t quite make sense of how it scores GPT 3.5, GPT 4, and GPT 4o. Going from 3.5 to 4 felt like a revolutionary step up to me, and I still haven&#x27;t quite grasped the quality improvement between 4 and 4o - 4o just seems a little chattier to me than 4. (My perceptions are through the web, not through the API.)
Tiberium10 months ago
Would be nice if they at least published the system prompts used (they can affect the performance of the models), and whether they use few-shot prompting or not.
评论 #41124809 未加载
attentive10 months ago
I wonder why their Groq llama-3.1-70b-versatile gets 81t&#x2F;sec when I get 250t&#x2F;sec on the same model.
xfalcox10 months ago
Is it just me or the 2 Llama 3.1 providers are the two know for running quantization levels that undermine Llama perf ?