PulseAugur
EN
LIVE 09:14:07

AI agent proves ARC-AGI-3 puzzles unwinnable, outperforming Claude and GPT

By PulseAugur Editorial · [1 sources] ·

Researchers from Tsinghua University and Zhejiang University have developed an Algebraic World Model (AWM) capable of solving ARC-AGI-3 puzzles. This agent, formalized in Lean 4, can also formally prove when a puzzle level is unwinnable. In testing, AWM proved all six sealed levels of ARC-AGI-3 to be unwinnable, a feat that Claude Opus 5.5 and GPT-6 Astra failed to achieve across 12 runs. AI

IMPACT Demonstrates novel AI reasoning capabilities and sets a new benchmark for AI problem-solving in abstract reasoning tasks.

RANK_REASON The cluster describes a new AI agent and its performance on a specific benchmark, which falls under research. [lever_c_demoted from research: ic=1 ai=1.0]

Read on Mastodon — fosstodon.org →

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →

AI agent proves ARC-AGI-3 puzzles unwinnable, outperforming Claude and GPT

How we ranked this

Signal score
5 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Tool
The cluster describes a new AI agent and its performance on a specific benchmark, which falls under research. [lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration
Single-source cluster
Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics
paper, model release
Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance
High
Clearly on-topic for AI-industry coverage.
Story freshness
Breaking (< 6h)
Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

Full methodology in our editorial standards.

COVERAGE [1]

  1. Mastodon — fosstodon.org TIER_1 English(EN) · [email protected] ·

    Tsinghua and Zhejiang University researchers have built an Algebraic World Model, an agent that plays ARC-AGI-3 puzzles and can formally prove a level cannot be

    Tsinghua and Zhejiang University researchers have built an Algebraic World Model, an agent that plays ARC-AGI-3 puzzles and can formally prove a level cannot be won. Formalised in Lean 4, it proved all six sealed levels unwinnable, where 12 runs of Claude Opus 5.5 and GPT-6 Astra…

    LINKS pandaily.com/tsinghua-algebraic-world-mod…

RELATED ENTITIES

RELATED TOPICS