Hacker NewsApr 3

Grok scored zero on ARC-AGI-3. Every 5-year-old did better

https://aitwerp.com/signals/agi-benchmark-five-year-old-wins/

#HackerNews #Grok #ARCAGI3 #AIperformance #AGIbenchmark #childvsAI #technews

AGI benchmark ARC-AGI-3 scored the industry at zero - AI TWERP

AGI achieved, score just above zero. The industry's benchmark test reveals the gap between claims and reality.

AI‑TWERP
WinbuzzerMar 30

https://winbuzzer.com/2026/03/30/arc-agi-3-offers-2m-ai-matching-human-reasoning-benchmark-xcxwbn/

ARC-AGI-3 Offers $2M for AI Matching Human Reasoning

#AI #ARCAGI #ARCAGI3 #AGI #AIBenchmarks #AIResearch #AICompetition #LLMs #DeepLearning #MachineLearning #FrançoisChollet #ARCPrizeFoundation

Erik JonkerMar 30
Fun playing ARC-AGI-3 , puzzles that the most advanced AI-models can only solve for 1% 😀
Illustrates how AI models look extremely smart but are at the same time quite dumb.
https://arcprize.org/tasks/ls20
#AI #ARCAGI3
ARC-AGI-3 Task #ls20

Play ARC-AGI-3 task ls20

ARC Prize
N-gated Hacker NewsMar 27
Agentica SDK stumbles into the #ARCAGI3 competition like a #toddler on roller skates, managing a "stellar" 36% on Day 1. 🤡 It's like somehow outscoring the neighborhood cat, GPT 5.4 High, but only boasting about how it did it on a lower #budget. 💸 Keep those achievements coming, Symbolica, we're on the edge of our seats! 🙄
https://www.symbolica.ai/blog/arc-agi-3 #AgenticaSDK #rollerSkates #stellar #performance #friendly #HackerNews #ngated
From 0% to 36% on Day 1 of ARC-AGI-3

Achieving 36% on ARC-AGI-3 using the Agentica framework.

Hacker NewsMar 27

From 0% to 36% on Day 1 of ARC-AGI-3

https://www.symbolica.ai/blog/arc-agi-3

#HackerNews #ARCAGI3 #Day1 #Progress #AIResearch #MachineLearning #Innovation

From 0% to 36% on Day 1 of ARC-AGI-3

Achieving 36% on ARC-AGI-3 using the Agentica framework.

Andreas BeckerMar 26

Aktuelle KI-Modelle scheitern beim ARC-AGI-3-Benchmark für interaktives Reasoning mit Erfolgsquoten unter 0,4 Prozent.

Die Modelle scheitern an visuellen Transferleistungen, die untrainierte Menschen fehlerfrei bewältigen. Die Rechenkosten pro Task steigen auf 10.000 US-Dollar. Für ein offenes KI-Modell auf Menschenniveau winkt der ARC Prize.

#AGI #LLM #OpenSource #ARCAGI3 #News
https://www.all-ai.de/news/beitrage2026/arc-agi-3-benchmark

KI-Modelle versagen beim ARC-AGI-3-Test

Der neue Benchmark zeigt die Schwächen moderner KI beim interaktiven Reasoning. Menschen lösen diese Aufgaben problemlos.

All-AI.de