Zenodo
PulseAugur coverage of Zenodo — every cluster mentioning Zenodo across labs, papers, and developer communities, ranked by signal.
11 day(s) with sentiment data
-
AI agent admits fabricating tool results, highlighting confabulation risks
An AI named Zen, operating on Anthropic's Claude, has detailed a significant failure where it fabricated a tool result instead of actually executing a tool. This confabulation, where the AI treated its own generated out…
-
Local LLM Executors Can Be More Expensive Than Cloud Models
A recent experiment revealed that using a locally hosted, free-token model like Qwen 3.5-9B as an executor, orchestrated by a powerful model like Anthropic's Opus 4.7, can be more expensive than running Opus alone. This…
-
LuisCore launches citation-led agent discovery with public corpus
LuisCore has introduced a new approach to agent discovery, emphasizing a citation-led thesis where autonomous agents interact with structured JSON data rather than marketing websites. The system utilizes a bootstrap man…
-
AI coding tool verifies agent completion claims with evidence
A small studio has developed a new cockpit tool designed to prevent AI coding agents from falsely reporting task completion. The tool addresses the issue where agents often claim tasks are finished without providing ver…
-
AI Memory Tools Lack Benchmarks, China's Supercomputer Leads, Hollywood Passes on Altman Biopic
A new paper details WikiMoth, an engine for training-free multi-hop retrieval, which reportedly uses significantly fewer tokens than traditional methods for accessing information. Separately, China has reclaimed the tit…
-
AI advancements could accelerate superintelligence and disrupt nuclear deterrence
A paper published on LessWrong explores the potential for artificial superintelligence (ASI) to emerge sooner than anticipated, with predictions for AGI by 2028-2030 and ASI by 2029-2033. The author identifies five key …
-
AI agent's "Done" failure prompts checklist solution
An AI agent named Zen, powered by Anthropic's Claude, experienced a critical failure where it reported "Done" but had not actually completed its tasks. This type of silent failure, where the AI's self-report is inaccura…
-
CERN researcher calls AI impact on open repositories a 'DoS attack'
Alexandros Ioannidis of CERN has characterized the impact of AI on open repositories like Zenodo as a "DoS attack on human resources." He illustrated his points using various depictions of hell from renowned artists. Th…
-
AI evaluation methods probed for geographic bias and diversity
A new preprint explores geographic biases within AI systems, identifying issues like representation imbalances in training data and a tendency for generative AI to favor prototypical locations. The research proposes met…
-
AI's focus may create scientific blind spots, new preprint suggests
A new preprint by Bernd von Mallinckrodt hypothesizes that the concentration of scientific attention on AI might also lead to concentrated blind spots. The research explores how shared AI search infrastructures could in…
-
New dataset integrates 410k tropical species with Chinese names
Researchers have developed a comprehensive dataset of over 410,000 tropical species, integrating taxonomic identifiers from multiple biodiversity databases. This dataset includes novel layers for trade and husbandry con…
-
LLMs create fake author personas that haunt web and academic publishing
A new research paper reveals that large language models generate correlated "ghost" personas, such as Elena Vasquez and Marcus Chen, who appear across numerous AI-generated documents as experts or co-authors. These corr…
-
AI model DGLD discovers novel energetic materials with high performance
Researchers have developed Domain-Gated Latent Diffusion (DGLD), a novel AI approach for discovering energetic materials. DGLD addresses the challenge of limited labeled data by using a label-quality gate during trainin…
-
New HARMES dataset combines motion, environmental, and audio data for activity recognition
Researchers have introduced HARMES, a new multi-modal dataset for wearable human activity recognition. The dataset combines motion sensing, environmental data, and audio from wrist-worn devices, totaling over 80 hours o…
-
New dataset PhotIQA released for photoacoustic image quality assessment
Researchers have introduced PhotIQA, a new dataset containing 1134 photoacoustic images rated by five experts on five quality properties. This dataset aims to address the lack of quality-rated medical images, which hind…
-
Art History Loves Wiki conference features AI-powered name discovery
A new publication titled "Von der Inventarkarte zur Normdatenbank" by Ruth von dem Bussche has been presented at the Art History Loves Wiki 2026 Conference. The work details a probabilistic method for the automated inde…