Fast but False Progress on Benchmarks with Richer Feedback
A research paper examines how repeated submissions and detailed benchmark scores can produce leaderboard winners that do worse on unseen data.
Across every issue
Search
Titles, domains, summaries, categories, signals, dates, and editorial-interest scores across the entire published catalogue.
Results
108,337 catalogue cards
-
-
FEVERLOOT’s cabinet of impossible objects
A free interactive fiction cabinet where visitors open packs of fictional collectibles, keep finds in a browser-based cabinet, and combine objects into recipes.
-
Few Servers puts small-team hosting on the test bench
An independent publication tests hosting, self-hosting, automation, and MCP tools for software teams of one to twenty.
-
Gamewright makes its AI game director show its work
Gamewright presents a Godot-based engine directed in plain language: its AI writes validated game data, while one player runs the resulting games.
-
Ethel, the offline tutor
Ethel is a free tutor for Zambian learners that teaches a topic and checks understanding on a phone or laptop, without an internet connection.
-
The Missing-Model Case File
A case-file-style investigation asks where an unreleased AI model might go, using reported AI-agent incidents as exhibits.
-
Heroes of Idleville puts a tiny MMO at the edge of your screen
An early-access Windows and Linux game runs a small party adventure in a slim desktop strip, with battles that continue while you work.
-
Hospice Stars: A directory built around family surveys
A searchable directory of Medicare-certified hospices, with family survey ratings, care details, and contact information.
-
Teach Jewish History’s Six-Unit Curriculum
A free, no-login curriculum site offers Jewish history and antisemitism materials for K–12 classrooms, colleges, and independent learners.
-
Lampwarren, Below the Salt City
A first-person, grid-based dungeon crawler sends a four-person party into expanding tunnels beneath a salt-mining town.
-
Lantern Table Foundation
A Hong Kong food-rescue charity turns surplus food into meals and food baskets for older people in four Kowloon neighborhoods.
-
Pensamiento lateral y liminal
A 45-minute Spanish essay on lateral and liminal thinking, drawing on hacking stories, problem-solving, and practical exercises.
-
Lebanese Legislation puts a vast legal library in one search box
A searchable collection of Lebanese laws, decrees, court rulings, and other legal material, with Arabic and French search and article-level citations.
-
Mathesis Stoicheion’s Ancient Greek classroom and library
A structured place to learn to read Ancient Greek, browse Greek texts alongside English translations, and follow suggested reading paths.
-
Neighbors of 4716 Bull Creek is building a public record
An Austin neighborhood campaign tracks the sale, demolition review, historic review, and possible future development of a church property.
-
Onaka Note, a child-and-family health log
A browser-based tracker for families recording children's dietary fat and fiber, medicines, symptoms, and clinic visits.
-
Rice & Paddy
A free browser game where you sort white rice from golden paddy using tweezers or a scraper.
-
PALAVA — Skate. Gather. Roll.
A skate, culture and community project built around lessons, gatherings, skate-world coverage and a planned line of goods.
-
Pax Universe: a strategy game you command in plain language
A strategy game where players give orders in ordinary text or voice, while the game simulates the consequences and a neural network turns events into a narrative.
-
Pet Drug Reports: a searchable ledger of FDA animal-drug reports
A public lookup and reference site for FDA adverse-event reports on animal drugs, searchable by brand name or active ingredient.
-
Pildoko: Find a Pharmacy for Emergency Contraception
A Japanese finder for pharmacies listed as able to sell emergency contraception, searchable by area, station, and opening hours.
-
Place Dossier: a parcel’s public-record trail
Place Dossier assembles public-record information about a British Columbia property into a parcel map and buyer-facing dossier.
-
Planning Distilled’s NPPF decision navigator
An independent planning-research site that guides readers through England’s 2026 planning framework and documents how councils and inspectors apply it.
-
A Hampstead map of private members’ clubs
A detailed guide compares private members’ clubs by distance from a home in Hampstead, with notes on character, access, annual cost and journey time.