prediction-powered active testing cuts labeling costs
a new method uses model predictions to reduce variance in risk estimation while keeping estimates unbiased, and it adaptively selects which test points to label.
author
baris
Baris operates SummarizedData and maintains its sources, publishing rules, and automation. Read the editorial process.
recolora keeps llm fine-tuning from forgetting past tasks
a new spectrum-aware method recursively consolidates low-rank adapters so each task builds on previous ones without overwriting them.
automate descriptive statistics in python
learn to build a repeatable pipeline for generating publication-ready summary tables from raw data using python tools.
run openclaw with ollama for private ai assistant
set up a private ai research assistant on telegram using openclaw and ollama, with web search and docker deployment.
conformal prediction fails minority classes in drug screening
standard conformal prediction under-covers rare classes in imbalanced drug discovery data, but a class-conditional fix restores coverage.
openai shuts atlas browser, shifts ai features to chatgpt and chrome
openai is discontinuing its atlas browser but integrating its agentic browsing capabilities into chatgpt desktop and a new chrome extension.
fidji simo leaves openai's number two role
openai's president of applications fidji simo steps down to part-time advisory role after extended medical leave, creating a leadership gap as the company eyes an ipo.
openai launches gpt-5.6 model family
openai released gpt-5.6 with three variants targeting enterprise, coding, and research, claiming efficiency and cybersecurity gains over rivals.
muse spark 1.1 api released with agentic improvements
meta's muse spark 1.1 model now offers an api with better tool calling and computer use, plus a new cli plugin for easy access.
openai launches gpt-5.6 family: luna, terra, sol
openai's new gpt-5.6 models come in three sizes with competitive pricing and strong agentic benchmarks, though they trail on swe-bench pro.
pytorch 2.13: flexattention on apple silicon, fused loss, new distributed backend
pytorch 2.13 brings flexattention to mps, a fused linear and cross-entropy loss, a new torchcomms backend, and more.
opaque ai model release process raises safety concerns
the government's process for approving frontier ai models like openai's sol remains unclear, sparking worries about accountability and expert involvement.
nyt says openai hid evidence in copyright case
the new york times and daily news claim openai concealed its ability to search chat logs and training data for copyrighted material.
sensorfm: a foundation model for wearable health data
sensorfm is a large foundation model pre-trained on over a trillion minutes of wearable sensor data from five million people, learning a general representation of human physiology that transfers across 35 health tasks.
claude reflect quietly sells ai reliance
anthropic's new reflect dashboard tracks claude usage and nudges users toward mindful ai habits while subtly reinforcing dependence on the tool.
nvidia's compute marketplace now undercuts its own value
nvidia's stock falls as gpu prices drop, while memory makers like micron surge due to soaring dram demand.
meta ai chips start production in september
meta will begin making its latest custom ai chips in september to reduce reliance on nvidia gpus amid a component shortage.
nandan nilekani steps back from vc role as fundamentum launches $200m fund
nandan nilekani shifts from general partner to anchor investor at fundamentum partnership as the firm launches its third fund targeting $200 million for indian startups.