Aphantasia challenges a centuries-old theory of abstract thought

Aphantasia, the inability to form mental images, poses a serious challenge to an influential theory of abstract thought in ...
GitHub

Universal Image Restoration via Internalized Chain-of-Thought Reasoning

Abstract: Image restoration seeks to recover high-quality images from degraded inputs but becomes highly ill-posed under complex, mixed degradations. While unified all-in-one models are common, their ...
the-decoder

Even the latest AI models make three systematic reasoning errors, ARC-AGI-3 analysis shows

The ARC Prize Foundation analyzed 160 game runs of OpenAI's GPT-5.5 and Anthropic's Opus 4.7 on the ARC-AGI-3 benchmark. The results reveal three systematic error ...
Healthline

Abstract Thinking: What It Is, Why We Need It, and When to Rein It In

Abstract thinking helps you find patterns in data and explain meaning through stories and ideas. It also supports humor, creativity, and problem-solving beyond what you can observe. Concrete thinking ...
Popular Mechanics

Scientists Found AI’s Fatal Flaw—The Most Advanced Models Are Failing Basic Logic Tests

Here’s what you’ll learn when you read this story: Large language models (LLMs) like ChatGPT show reasoning errors across many domains. Identifying vulnerabilities is good for public safety, industry, ...
VentureBeat

Large reasoning models almost certainly can think

Recently, there has been a lot of hullabaloo about the idea that large reasoning models (LRM) are unable to think. This is mostly due to a research article published by Apple, "The Illusion of ...
GitHub

Training Vision-Language Process Reward Models (VL-PRMs) for Test-Time Scaling in Multimodal Reasoning

Pairing VL-PRMs trained with abstract reasoning problems results in strong generalization and reasoning performance improvements when used with strong vision-language models in test-time scaling ...
Ars Technica

LLMs’ “simulated reasoning” abilities are a “brittle mirage,” researchers find

In recent months, the AI industry has started moving toward so-called simulated reasoning models that use a “chain of thought” process to work through tricky problems in multiple logical steps. At the ...
VentureBeat

New AI architecture delivers 100x faster reasoning than LLMs with just 1,000 training examples

Singapore-based AI startup Sapient Intelligence has developed a new AI architecture that can match, and in some cases vastly outperform, large language models (LLMs) on complex reasoning tasks, all ...
Science Daily

Thinking AI models emit 50x more CO2—and often for nothing

Every query typed into a large language model (LLM), such as ChatGPT, requires energy and produces CO2 emissions. Emissions, however, depend on the model, the subject matter, and the user. Researchers ...
Frontiers

Some of your AI prompts could cause 50 times more CO2 emissions than others

Every query typed into a large language model (LLM), such as ChatGPT, requires energy and produces CO 2 emissions. Emissions, however, depend on the model, the subject matter, and the user.