Kimi K3 Outperforms Fable 5 in Frontend Code
Moonshot's Kimi K3 scored 1,679 in frontend code benchmark, surpassing Fable 5's 1,631, according to the Code Arena: Frontend test.
Research
190 articles
In-depth coverage of new AI research — papers, benchmarks, and breakthroughs from leading labs and academia, summarized for fast reading and grounded in the methods that matter.
LATEST NEWS
-
-
Meta and Stanford Test AI with Baby-Like Learning
Researchers developed EgoBabyVLM, a test that challenges AI models to learn like infants, using footage from babies' cameras. The test reveals current models struggle to interpret messy, real-world data.
-
ELIZA Source Code Reveals Chatbot's Multiple Personalities
Researchers uncovered the original ELIZA source code from MIT archives, showing the chatbot could assume various personas beyond its therapist role.
-
Hugging Face Evaluates Open-Source AI Models for Swiss Legal Tasks
Hugging Face evaluated 13 open models on three Swiss legal benchmarks with 55,861 scored samples each, showing the top five models differ by only 2.1 points.
-
Anthropic Discovers New Internal Space in AI Models
Anthropic's new research reveals a hidden 'J-space' within AI models that influences their reasoning, offering deeper insight into how they process information.
-
AMD Optimizes Video Sparse Attention on ROCm
AMD announces ROCm-optimized video sparse attention with 3.13× speedup over prior methods, reducing training-time latency to 73.06 ms.
-
Anthropic's J-Space Explores Consciousness in Claude
Anthropic's new research suggests Claude may have two distinct language processing routes, with one resembling access consciousness. The study shows the model can continue Spanish text while naming French authors, indicating potential for explicit reasoning.
-
Quantum Computing Boosts AI Drug Discovery Models
A Danish team used quantum computing to enhance AI models, generating more effective peptides for vaccine development.
-
Cohere Introduces Hardware-Aware Dynamic Speculative Decoding
Cohere's new technique improves large language model inference speed by adapting to hardware constraints, achieving up to 23% faster performance at high batch sizes.
-
Anthropic Unveils J-Space Inside Claude Opus 4.6
Anthropic has identified a hidden area within its Claude Opus 4.6 model called J-space, offering new insights into how the AI processes information.
-
OpenAI Finds 30 Percent of Popular AI Coding Test Broken
OpenAI has pulled its endorsement of SWE-Bench Pro after finding 30 percent of its tasks flawed, impacting AI performance assessments.
-
IBM Introduces CofrGenets to Improve Transformer Model Training
IBM researchers have developed CofrGenets, a new framework that reduces training errors in transformer-based models by 40%.