Subscription platforms (music, video, news) let users pay once for unlimited content. But how should that money be split among creators—without inviting fraud? New research by Abheek Ghosh, Tzeh Yuan Neoh, Nicholas Teh, and Giannis Tyrovolas studies revenue-sharing rules that are manipulation-resistant by design, not just guarded by ever-shifting
AI
Teach your AI judge—faster—with synthetic test cases Evaluating AI with "LLM-as-a-judge" works—until you run out of good examples. This study introduces a tool that generates rich, customizable synthetic test cases to refine your evaluation criteria, all within a human-in-the-loop workflow. * Configure domains, personas, lengths, desired
AI web agents can browse sites, fill forms, and compare prices for you—from OpenAI’s Operator to Google’s Project Mariner. But how green are they? * The study estimates and benchmarks the energy and CO2 cost of web agents. * Design choices (like model size and browsing strategy) can swing
AI
Do AI chatbots have personalities? Using the Big Five psychology test (BFI‑2), researchers probed six large language models and turned the temperature dial—the setting that makes replies more or less random. * Four of the five traits differed across models. * Neuroticism and Extraversion rose or fell as temperature changed.
AI
How human are today’s AI models when making social decisions? This study builds a "digital twin" of classic game-theory experiments and checks whether large language models (LLMs) act like people. * Llama closely replicates human cooperation patterns, including the ways people deviate from strict rational-choice rules. * Qwen behaves
Can we make AI tools in criminal justice both fair and accurate? A new paper by Shaolong Wu, James Blume, and Geshi Yeung says yes—with a practical tweak. Instead of forcing exact equality between groups (which can hurt accuracy or be impossible), they optimize overall errors while keeping differences
CivicTech
In citizens’ assemblies and online town halls, many people submit questions for experts—but only a few make it on stage. How do we choose a small set that truly reflects everyone’s concerns? Researchers introduce a way to audit representation of the chosen questions, grounded in a social-choice idea
AI
MusRec is a new AI model that edits existing songs using plain text—without task‑specific retraining. Powered by rectified flow and diffusion transformers, it brings true zero‑shot, text‑guided edits to real‑world music (not just audio generated by the model itself). * What it does: Apply diverse edits
LLM
Speed now, debt later? What a study of Cursor finds Do AI coding agents really make teams more productive? This large-scale study estimates the causal impact of adopting Cursor across GitHub projects using a modern difference-in-differences design. * Fast start: Projects see a significant, large—but short-lived—jump in development velocity.
AI
Tell richer stories, one node at a time This research introduces a node-based editor that lets you build a multimodal story as a graph—each “node” can hold text, images, audio, and video. You can expand, reorder, and refine nodes using plain-language prompts or direct edits. * Targeted edits: tweak a
AI
Tell richer stories, one node at a time This research introduces a node-based editor that lets you build a multimodal story as a graph—each “node” can hold text, images, audio, and video. You can expand, reorder, and refine nodes using plain-language prompts or direct edits. * Targeted edits: tweak a
AI
Tell richer stories, one node at a time This research introduces a node-based editor that lets you build a multimodal story as a graph—each “node” can hold text, images, audio, and video. You can expand, reorder, and refine nodes using plain-language prompts or direct edits. * Targeted edits: tweak a