#llmbenchmark — Public Fediverse posts

Antigravity 2.0 Tops the OpenSCAD Architectural 3D LLM Benchmark

https://modelrift.com/blog/openscad-llm-benchmark/

#HackerNews #Antigravity #2.0 #OpenSCAD #3DPrinting #LLMBenchmark #Architecture

That's what I call a «meaningful» LLM benchmark. 😉

(... or how to debunk the German meaning of «Intelligenz».)

https://petergpt.github.io/bullshit-benchmark/viewer/index.v2.html

#ai #llm #llmbenchmark #benchmark

Google’s new Gemini 3.1 Pro claims to double its reasoning scores on the latest benchmark, pushing LLM capabilities further. Curious how this stacks up against other open‑source models? Dive into the details and see what the numbers reveal. #GoogleGemini #Gemini3_1 #LLMbenchmark #GenerativeAI

🔗 https://aidailypost.com/news/google-gemini-31-pro-doubles-reasoning-performance-benchmark

New benchmark shows top LLMs struggle in real mental health care

https://swordhealth.com/newsroom/sword-introduces-mindeval

#HackerNews #LLMbenchmark #MentalHealth #AIinHealthcare #MentalHealthTech #HealthcareInnovation

The proof that #benchmarks on #LLM models are utterly useless.

Maybe it's time to focus on real-world performance and practical applications instead of chasing numbers?

#llm #ai #aibenchmarks #llmbenchmark #machinelearning #artificialintelligence #openai #gpt5 #chatgpt

🚀 Featured in L'Usine Digitale!

Our independent multilingual LLM benchmark Phare was highlighted in an article detailing some key insights from our research.

🔎 Key finding: LLMs perpetuate biases in their own content while recognizing those same biases when asked directly.

Thanks to L'Usine Digitale and Célia Séramour for this coverage.
Read here: https://gisk.ar/4lCHoUB

#LLMBenchmark #AISafety #AISecurity

🚀 Claude 4 didn’t just assist—it outperformed.
In a 7-hour live dev session, it refactored legacy Java with zero hallucinations, full memory, and enterprise-grade precision.

🔍 We compared Claude 4 vs ChatGPT across 5 key metrics — and the results will surprise you.

📖 Read the full breakdown:
👉 https://medium.com/@rogt.x1997/claude-4-vs-chatgpt-the-5-metrics-that-prove-its-not-just-an-assistant-48f82384c69f

📌 #Claude4 #LLMbenchmark #AIengineering #Anthropic
https://medium.com/@rogt.x1997/claude-4-vs-chatgpt-the-5-metrics-that-prove-its-not-just-an-assistant-48f82384c69f

Thanks to Kyle Wiggers for this article. We're honored to see our research covered by TechCrunch. 🤝

Read the article here: https://techcrunch.com/2025/05/08/asking-chatbots-for-short-answers-can-increase-hallucinations-study-finds/

#AISecurity #LLMBenchmark #research

The article present some key findings from our benchmark:
- Most widely used models aren't necessarily the most reliable
- Some models tend to agree with users regardless of factual accuracy
- The way questions are phrased impacts response reliability

Thanks to Les Echos and Joséphine Boone for this coverage 🤝

Read the article here: https://www.lesechos.fr/tech-medias/intelligence-artificielle/desinformation-rumeurs-influences-quelles-ia-hallucinent-le-plus-2163628

#AISecurity #LLMBenchmark #LesEchos

Phare is developed by Giskard with Google DeepMind, the European Commission and Bpifrance as research & funding partners.

👉 Full analysis: https://www.giskard.ai/knowledge/good-answers-are-not-necessarily-factual-answers-an-analysis-of-hallucination-in-leading-llms
Benchmark results: https://phare.giskard.ai

#AISecurity #LLMBenchmark #LLMs

The replay of our session at Forum INCYBER Europe (FIC) is now online 🎬

Watch our CTO present the initial Phare results - our multilingual and independent LLM benchmark that evaluates hallucination, factual accuracy, bias, and harm potential.

The session features Matteo Dora and Elie Bursztein (Google DeepMind).

Full recording linked below 👇

#LLMBenchmark #AISecurity #ForumINCYBER #Research

✨ Announcing Phare: new multi-lingual #LLMBenchmark 🌊

We're announcing an open & independent LLM benchmark to evaluate key AI security dimensions including hallucination, factual accuracy, bias, and potential for harm across several languages, with @googledeepmind as research partner.

Phare (Potential Harm Assessment & Risk Evaluation) will cover leading models from the top 7 AI labs in English, French, and Spanish, and will evaluate models across four dimensions:
👇