synthetic-data
WROP trains video world models to keep track of hidden objects News
A new controlled training corpus and fine-tuned video model improve object permanence in short continuations, while leaving collision and general long-horizon world modeling unresolved.
Quasar 1.1 used quantum-generated data, not quantum training News
Multiverse Computing says an IBM Heron helped generate a small synthetic-data component for healing a compressed 438B classical model, without showing that quantum hardware caused its reported gains.
Pseudo-labeling and self-training Lesson
Pseudo-labeling is training a model on labels it produced itself: run a model over unlabeled data, keep the predictions it is confident about, and treat them as ground truth for the next round of training. It works surprisingly well, and it fails in one specific way -- by confidently reinforcing its own mistakes.
Ornith-1.5 writes its own training problems and grades them News
Ornith released an open-weight model family whose training loop generates its own tasks, builds its own scoring harnesses, and feeds the reward back into all three stages -- with the 397-billion-parameter flagship matching Claude Opus 4.8 on agentic coding benchmarks.
Gallup is testing AI agents that answer surveys for real people News
Gallup has built AI agents from in-depth interviews with about 1,000 of its panel members and is independently validating whether their simulated answers can stand in for human ones, while its partner Simile raised over $200 million at a $2 billion valuation.
Three papers landed the same day arguing you should build the world, not the model News
EnvHarness, FACET, and SPADE all took the top spots on Hugging Face's daily paper list with the same underlying move, shifting effort from making the agent smarter to manufacturing the environments the agent practices in, with FACET releasing 6,020 ready-made terminal tasks.
A task factory ran fifteen rounds and broke the model grading it News
A new paper builds harder and harder terminal tasks by recursively rewriting accepted ones, and across fifteen rounds a fixed frontier solver's success rate fell from 90 percent to 2.5 percent, with the authors reporting no ceiling in sight.
Synthetic Data: When AI Makes Its Own Training Material Lesson
The internet is running out of fresh text to train on, so the most advanced models increasingly learn from data that other AI made or shaped. Here is how that works, why it helps, and how it can quietly poison a model.
Recursive-Task-Synthesis Tool
A public set of 37,484 verified long-horizon terminal-agent tasks, each a runnable bundle with instruction, environment, reference solution and hidden verifier, plus a companion set of 327,000 agent trajectories and three fine-tuned Qwen3.5 checkpoints.