Pearson
PulseAugur coverage of Pearson — every cluster mentioning Pearson across labs, papers, and developer communities, ranked by signal.
6 day(s) with sentiment data
-
New AI evaluation suite CrisisBench targets crisis counseling capabilities
Researchers have developed CrisisBench, a new evaluation suite designed to assess the capabilities of AI models in crisis counseling scenarios. The suite includes over 1,000 tasks derived from de-identified crisis conve…
-
Australian universities use AI for marking, raising concerns over assessment integrity
Australian universities are increasingly using AI for grading, with Pearson's systems having processed 39 million responses since 1998. A joint study from Melbourne and Cardiff universities found that ChatGPT's grading …
-
AI framework simulates urban mobility patterns in Niterói, Brazil
Researchers have developed a computational framework to simulate urban mobility patterns using aggregated urban data. This framework combines synthetic populations, the built environment, and transport networks to gener…
-
New framework predicts AI loss-of-control risk with 0.84 correlation
Researchers have developed a new behavioral framework to predict and understand the risk of loss of control in frontier AI systems. This framework breaks down the risk into three factors: misaligned motive, harm-enablin…
-
Anthropic launches Claude Certified Developer Exam for API users
Anthropic has launched a "Claude Certified Developer Exam: Foundations" for engineers working with its Claude API and related tools. The exam, which costs $125 and is currently restricted to individuals within the Claud…
-
Textbook authors protest unfair settlement process and publisher splits
Textbook authors, represented by Kim Pawlak of the Textbook and Academic Authors Association, have expressed dissatisfaction with a settlement process. They are complaining about a confusing claims procedure and what th…
-
New DRAFT method reconstructs 3D room layouts from fMRI and camera data
Researchers have developed a novel method called DRAFT that can reconstruct the 3D layout of rooms using only functional magnetic resonance imaging (fMRI) signals and camera trajectories, without relying on visual frame…
-
AI agents accelerate molecular discovery through iterative lab feedback
Researchers have demonstrated a novel approach to molecular discovery by integrating AI agents with wet-lab experiments in a closed-loop system. This method, tested using protein-binding aptamer design against alpha-thr…
-
Language models' generalization ability linked to linear fact readability
Researchers have identified a key factor in how well language models generalize knowledge after fine-tuning. They distinguish between "active facts," which aid generalization, and "inert facts," which do not. The study …
-
New benchmark Game2Real-Bench shows toy games correlate with real-world VLM reasoning
Researchers have introduced Game2Real-Bench, a new benchmark designed to assess the visual reasoning capabilities of vision-language models (VLMs). This benchmark aims to determine if skills learned in toy games can tra…
-
New Persona Hierarchy Model Explains LLM Generalization
Researchers have introduced the Persona Hierarchy Model to explain why fine-tuned language models sometimes generalize broadly and other times remain context-specific. The model posits that a shared default persona infl…
-
AI security threats rise: LLMjacking, upskilling needs, and new iPhone scam protection
Google analysts are warning about a surge in "LLMjacking," where stolen AI credentials are sold on the black market, leading to increased costs for businesses. Separately, Pearson is acquiring Workera to help companies …
-
X's recommender system inadvertently profiles user political ideology
A new research paper published on arXiv details how X's "Who to Follow" recommendation system inadvertently profiles users' ideological positions. Researchers analyzed recommendations shown to volunteers in France and r…
-
Inspectable Math Verification Contract Built in Pythos in 24 Hours
This article details the process of building an inspectable math verification contract in Pythos, a Python-based language, within a 24-hour timeframe. It highlights the journey from a comment on DEV Community to a funct…
-
RankShift algorithm detects categorical data shifts within databases
Researchers have developed RankShift, a novel method for detecting and explaining categorical shifts within analytical databases. This technique identifies changes in category distribution without altering the overall e…
-
New DMT-Dens method visualizes biological data while preserving density
Researchers have developed DMT-Dens, a new method for visualizing biological data in low dimensions. This technique uses a Transformer encoder to preserve the density of observations, which is crucial for accurately int…
-
New research models human detection of AI image edits in two stages
A new research paper proposes a two-stage model for understanding how humans detect AI-generated image edits, differentiating between attention capture and judgment accuracy. The study, involving 59 participants and an …
-
New frameworks enhance LLM temporal reasoning evaluation
Researchers have developed new frameworks to better evaluate the temporal reasoning capabilities of Large Reasoning Models (LRMs). One approach, TRACE, models temporal reasoning as constraint satisfaction problems using…
-
Foundation models adapted for cardiac MRI analysis achieve strong segmentation and ejection fraction estimation
Researchers have adapted foundation models for analyzing cardiac MRI scans, achieving strong results in segmenting different views of cine MRI and estimating ejection fraction. The models demonstrated high Dice scores f…
-
SoftMCC framework enhances model selection for imbalanced classification
Researchers have introduced SoftMCC, a novel post-training framework designed to improve model selection for imbalanced binary classification tasks. This method addresses the threshold-dependency issues inherent in trad…