All things data. One authority.

Topics

Data quality, labels and provenance

Good data is made with care. These stories follow the labels, checks and records of origin that make data worth training on.

Stories

Life sciences · Brief

Fake patient notes that doctors can often spot

A public set of synthetic Portuguese clinical notes helps research without exposing real patients, but it is not perfect.

19 November 2026 · 1 min read
AI and training data · Brief

When three readers disagree, the label hides it

A new Arabic dataset keeps every reader's verdict, and shows what a single majority label throws away.

19 November 2026 · 1 min read
A single wheat field divided by a hedge, half golden and half green, under an even skyPolicy and law
Policy and law · Brief

Same farms, five data sets, five different answers

Which data set a regulator picks can change which crops are told to cut their fertiliser.

16 November 2026 · 2 min read
A glass microscope slide lying on a white laboratory bench beside a pencilLife sciences
Life sciences · Brief

Authors pull their own mouth-cancer data paper

A 37-author team withdrew a dataset paper after finding bugs in its script, saying the reported figures may mislead.

13 November 2026 · 2 min read
A single branching hard coral on a sunlit shallow reef, seen from just aboveAI and training data
AI and training data · Brief

Even coral experts disagree on about one label in twelve

A reef dataset built from 76 scattered sources shows what it takes to make labelled images fit for training.

12 November 2026 · 2 min read
A single folded newspaper on a stone doorstep in morning lightAI and training data
AI and training data · Brief

Fourteen experts label bias in 448 Hindi news articles

A small open data set shows how a language left out of the tools gets its first measuring stick.

11 November 2026 · 1 min read
A brass balance scale with two unequal pans on a plain tableAI and training data
AI and training data · Brief

The most accurate post moderator is often not the fairest

A new test set shows that judging posts well and judging them fairly are different skills.

11 November 2026 · 1 min read
A single closed wooden door with a small brass question-shaped keyhole in a quiet corridorTechnology
Technology · Brief

Teaching an AI model to say it does not know

A new method nudges a few internal switches so a model declines to guess.

9 November 2026 · 2 min read
a hand-drawn paper maze with one wall broken openAI and training data
AI and training data · Brief

Machine-written tests catch more AI failures than human ones

A method that grounds test questions in real-world evidence makes chatbots fail more often on medical advice and self-harm.

7 November 2026 · 2 min read
a row of milk bottles on a doorstep, a few with fresher datesData value
Data value · Brief

Not every stale record is worth the cost of re-checking

A short paper shows that choosing which data to refresh by the task at hand beats refreshing blindly.

7 November 2026 · 2 min read
A brass balance scale with uneven stacks of small stones on a plain stone ledgeAI and training data
AI and training data · Brief

Reweighting training data can shift harm to another group

A preprint finds a common fairness fix lowers missed cases for one group by raising them for another.

6 November 2026 · 1 min read
A sealed glass specimen jar with a handwritten paper label on a laboratory shelfData value
Data value · Brief

Image checks cannot prove where a medical scan came from

A test of synthetic chest X-rays at dataset intake says provenance has to be recorded, not guessed from pixels.

5 November 2026 · 1 min read
a tidy wall of identical labelled glass specimen jars on a laboratory shelfLife sciences
Life sciences · Brief

A curated set and shared test make protein-model building comparable

CryoAtlas filters public structure records into one training set with a fixed benchmark.

4 November 2026 · 1 min read
A stack of old library index cards fanned across a wooden desk in morning lightAI and training data
AI and training data · Brief

A text-reuse test set built only from open sources

Researchers release a bilingual set for testing plagiarism detectors, with every source passage openly licensed or in the public domain.

4 November 2026 · 2 min read
A single tall glass jar of water on a laboratory bench, clear on one side and cloudy on the otherLife sciences
Life sciences · Brief

A tool that cleans health records for AI is put to the test

A new benchmark checks a chat agent that turns raw patient records into datasets ready for AI, and reports where it falls short.

4 November 2026 · 2 min read
a single unlabelled brown glass medicine bottle standing on a white laboratory benchLife sciences
Life sciences · Brief

A public trial registry trains a simulator that ignores the drug

A new test shows models built on registry data can score well without using the one field that matters.

4 November 2026 · 1 min read
a long row of weathered paper flight logbooks stacked on a hangar shelfTechnology
Technology · Brief

One aviation safety dataset built from 128 sources

A new paper merges national accident records into one deduplicated, traceable corpus.

3 November 2026 · 2 min read
a stack of copper plates in a laboratory, edge on, with soft light along the metalPhysics and frontier science
Physics and frontier science · Brief

A detector's full record can tell more than the single number it is usually boiled down to

A new preprint finds that using the whole simulated record of a particle shower sharpens the energy reading by a factor of 1.7 to 1.9.

30 October 2026 · 1 min read

Get the weekly issue

One email a week once the issue starts, only the desks you pick. Leave whenever you like.

Your desks
Or follow topics
Your topics

We use your email only to send the issue. Your rights.

The data desks

Value
Protect
Share
Grow

Desks marked ↗ lead to related partner sites.