# Data Generation

Ai2 released ToyBlocks, an open simulation framework built around a robot arm and blocks on a tabletop and the first of its FoundationWorlds. It automatically generates tasks and demonstrations and reports metrics beyond success rates, such as execution speed, forces exerted and task progress, so researchers can change training data or model design, repeat a test and trace a failure. In zero-shot real-robot tests, ManiFlow's atomic-skill success rates were 80.0% in simulation and 81.1% in reality.

Ai2 released ToyBlocks, an open simulation framework built around a robot arm and blocks on a tabletop and the first of its FoundationWorlds. It automatically generates tasks and demonstrations and reports metrics beyond success rates, such as execution speed, forces exerted and task progress, so researchers can change training data or model design, repeat a test and trace a failure. In zero-shot real-robot tests, ManiFlow's atomic-skill success rates were 80.0% in simulation and 81.1% in reality.

ELEM Biotech plans to invest US$1 million in generating and openly sharing perturbation data for training and validating virtual cell models, and is asking researchers worldwide which experiments should be measured first. Submissions must name concrete cell, gene and perturbation combinations along with sampling times, readouts, controls and independent biological replicates, building on the already open-source ELEM-1200RS and the upcoming ELEM-600PR. Its automated platform can prepare up to 100,000 arrayed genetically perturbed cell samples per month, but that is sample production capacity, not 100,000 distinct genes measured monthly.

ELEM Biotech plans to invest US$1 million in generating and openly sharing perturbation data for training and validating virtual cell models, and is asking researchers worldwide which experiments should be measured first. Submissions must name concrete cell, gene and perturbation combinations along with sampling times, readouts, controls and independent biological replicates, building on the already open-source ELEM-1200RS and the upcoming ELEM-600PR. Its automated platform can prepare up to 100,000 arrayed genetically perturbed cell samples per month, but that is sample production capacity, not 100,000 distinct genes measured monthly.

Two-stage stochastic service network design must fix how many vehicles run on each arc before demand is known, and the recourse cost of adding spot vehicles makes the problem hard to solve. A feedforward neural network replaces that recourse subproblem and is embedded in a mixed-integer program; because first-stage variables are general integers, training samples come from perturbing solutions of single-scenario problems. Across 18 instances, designs from the perturbation approach deviate 1.33% on average from their full-scenario evaluated cost, versus 10.81% for random sampling.

Two-stage stochastic service network design must fix how many vehicles run on each arc before demand is known, and the recourse cost of adding spot vehicles makes the problem hard to solve. A feedforward neural network replaces that recourse subproblem and is embedded in a mixed-integer program; because first-stage variables are general integers, training samples come from perturbing solutions of single-scenario problems. Across 18 instances, designs from the perturbation approach deviate 1.33% on average from their full-scenario evaluated cost, versus 10.81% for random sampling.

Treating one successful demonstration as an editable skeleton, a learned editor decides when to switch between object-centric stages and how to correct each stage geometrically from sparse success rewards, reaching 92.7% average success on 15 Meta-World tasks and 98.4% average edit success on 9 DexMimicGen tasks.

Treating one successful demonstration as an editable skeleton, a learned editor decides when to switch between object-centric stages and how to correct each stage geometrically from sparse success rewards, reaching 92.7% average success on 15 Meta-World tasks and 98.4% average edit success on 9 DexMimicGen tasks.

The work presents a meta-agent pipeline in which a frontier model such as Claude Opus generates terminal tasks and verifiers, diagnosing three failure classes—benchmark invalidity, harness brittleness, and reward misalignment—and reports that prompt redesign and context extension raise baseline solvability 5.6 times, that a 9B model saturates at 81.3% mean pass@2 within 20 steps on Claude Opus-generated tasks, and that adding hard tasks reduces mean pass@2 to 20.6% without changing the training configuration, indicating that the solvability band is model-specific.

The work presents a meta-agent pipeline in which a frontier model such as Claude Opus generates terminal tasks and verifiers, diagnosing three failure classes—benchmark invalidity, harness brittleness, and reward misalignment—and reports that prompt redesign and context extension raise baseline solvability 5.6 times, that a 9B model saturates at 81.3% mean pass@2 within 20 steps on Claude Opus-generated tasks, and that adding hard tasks reduces mean pass@2 to 20.6% without changing the training configuration, indicating that the solvability band is model-specific.