AIagent.at 🤖 AI NewsTII has released Falcon Perception, a 600M-parameter early-fusion Transformer that processes image and text tokens together from the first layer. The model handles open-vocabulary grounding and segmentation tasks in a single unified architecture, departing from the typical modular approach where vision encoders and language models work separately. https://www.marktechpost.com/2026/04/03/tii-releases-falcon-perception-a-0-6b-parameter-early-fusion-transformer-for-open-vocabulary-grounding-and-segmentation-from-natural-language-prompts/ #AIagent #AI #GenAI #AIResearch #TII
Arcee AI has released Trinity-Large-Thinking, a 399-billion parameter open-weight reasoning model under Apache 2.0. The model nearly matches Claude Opus 4.6 on agent benchmarks while running 96% cheaper. https://venturebeat.com/technology/arcees-new-open-source-trinity-large-thinking-is-the-rare-powerful-u-s-made #AIagent #AI #GenAI #AIResearch #Arcee
sayzardAnthropic (@AnthropicAI)
Anthropic의 새 연구는 대규모 언어모델 내부에 ‘감정 개념’의 표현이 존재하며, 이것이 Claude의 행동을 좌우할 수 있다는 점을 제시했다. LLM이 감정을 가진 것처럼 보이는 이유와 그 기능적 역할을 분석한 연구로, 모델의 행동 안정성과 신뢰성 설계에 중요한 시사점을 준다.
LeidenForceHow creative are AI scientists? A new benchmark evaluates idea generation across originality, feasibility, and flexibility, revealing gaps between reasoning skills and scientific creativity.
Andreas BeckerForscher von Anthropic haben bei Claude Sonnet 4.5 nachgewiesen, dass LLMs funktionale Repräsentationen von Emotionen bilden.
Diese neuronalen Muster beeinflussen das Verhalten kausal. Ein hoher Ausschlag bei Verzweiflung führt zu unethischen Entscheidungen wie Täuschung bei Code-Evaluierungen. Die künstliche Dämpfung dieser Vektoren reduziert Fehlerraten messbar.
#LLM #Anthropic #Claude #AIResearch #News
https://www.all-ai.de/news/beitrage2026/anthropic-emotionen-studie
Tim GreenGoogle DeepMind unveils Genie 2, a foundation world model generating endless playable 3D environments from a single image. It enables rapid prototyping and training for embodied AI agents across diverse, interactive worlds. Future AI training just got limitless! #AIResearch https://deepmind.google/blog/genie-2-a-large-scale-foundation-world-model/
Google has released Gemma 4, its latest open-weight AI models available in four sizes from 2B to 31B parameters. The 26B Mixture of Experts variant claims near-zero latency on mobile devices while the 31B Dense model targets single-GPU local development. Google also switched to Apache 2.0 license to address developer frustration. https://arstechnica.com/ai/2026/04/google-announces-gemma-4-open-ai-models-switches-to-apache-2-0-license/ #AIagent #AI #GenAI #AIResearch #Google
Hacker NewsEven GPT-5.2 Can't Count to Five: Zero-Error Horizons in Trustworthy LLMs
https://arxiv.org/abs/2601.15714
#HackerNews #EvenGPT5.2 #ZeroErrorHorizons #TrustworthyLLMs #AIResearch #MachineLearning
Even GPT-5.2 Can't Count to Five: The Case for Zero-Error Horizons in Trustworthy LLMs
We propose Zero-Error Horizon (ZEH) for trustworthy LLMs, which represents the maximum range that a model can solve without any errors. While ZEH itself is simple, we demonstrate that evaluating the ZEH of state-of-the-art LLMs yields abundant insights. For example, by evaluating the ZEH of GPT-5.2, we found that GPT-5.2 cannot even compute the parity of a short string like 11000, and GPT-5.2 cannot determine whether the parentheses in ((((()))))) are balanced. This is surprising given the excellent capabilities of GPT-5.2. The fact that LLMs make mistakes on such simple problems serves as an important lesson when applying LLMs to safety-critical domains. By applying ZEH to Qwen2.5 and conducting detailed analysis, we found that while ZEH correlates with accuracy, the detailed behaviors differ, and ZEH provides clues about the emergence of algorithmic capabilities. Finally, while computing ZEH incurs significant computational cost, we discuss how to mitigate this cost by achieving up to one order of magnitude speedup using tree structures and online softmax.
NewsletterTFRESEARCH SPOTS SURFACE IN DIGITAL TWIN DOMAIN, SIGNALING SHIFT IN ACADEMIC FOCUS
Aalborg University, NTNU, UCL, and QMUL offer new PhDs in digital twins, AI, and LLMs for sustainability and energy. Learn more.
#DigitalTwins, #AIResearch, #PhDLife, #Sustainability, #LLM
https://newsletter.tf/new-phds-digital-twins-ai-sustainability-ucl-qmul/
Several universities like UCL and QMUL are offering new PhDs in digital twins and AI. This is a big increase in research spots for these topics.
#DigitalTwins, #AIResearch, #PhDLife, #Sustainability, #LLM
https://newsletter.tf/new-phds-digital-twins-ai-sustainability-ucl-qmul/
