Data Curation & Synthetic Data - Weekly Roundup

Purely synthetic training data outperforms native real-world supervision on satellite stereo benchmarks, showing that pristine geometric labels and zero temporal noise beat noisy real-world data collection.

Collecting more errors doesn't guarantee better models: targeting only the high-impact failure modes signaled by a network's own internal representations outperforms standard rarity-based active learning while saving annotation budget.

Static problem pools cause LLM reasoning post-training to stall as models saturate the dataset, but dynamically synthesizing frontier tasks via regret-guided procedural generation keeps learning signals saturated for sustained gains.

Even when provided with only target column metadata, LLMs outperform specialized AutoML heuristics like AutoGluon at inferring machine learning task types (0.98 vs. 0.93 macro F1), unlocking fully autonomous pipeline configuration.

Contemporary frontier speech models still catastrophically fail on living Arabic dialects, with top systems like GPT-4o achieving an error rate no better than 35.0% WER across 17 regional vernaculars.

Collecting more errors doesn't guarantee better models: targeting only the high-impact failure modes signaled by a network's own internal representations outperforms standard rarity-based active learning while saving annotation budget.

Static problem pools cause LLM reasoning post-training to stall as models saturate the dataset, but dynamically synthesizing frontier tasks via regret-guided procedural generation keeps learning signals saturated for sustained gains.

Even when provided with only target column metadata, LLMs outperform specialized AutoML heuristics like AutoGluon at inferring machine learning task types (0.98 vs. 0.93 macro F1), unlocking fully autonomous pipeline configuration.

Contemporary frontier speech models still catastrophically fail on living Arabic dialects, with top systems like GPT-4o achieving an error rate no better than 35.0% WER across 17 regional vernaculars.

Jev’s apparent miscalibration on binary choices is an artifact of forced binarization: recovering a hidden, unexpressed "I don't know" probability mass catapults the model's calibration accuracy from 0.771 to 0.978.

Verifying LLM span extractions no longer requires full manual re-annotation: character-level policy reconciliation isolates genuine disagreements, slashing human review to just 5.4 focused conflicts per document.

Industrial anomaly detection usually breaks down where data is sparsest, but baking reaction kinetics directly into conditional diffusion enables high-fidelity trajectory synthesis of catastrophic failure states far outside the training distribution.

Model collapse in generative self-training can be weaponized: intentionally amplifying a generator's Jacobian-derived geometric distortions creates a far more potent negative guidance signal than standard synthetic outputs.

High-fidelity 3D texture generation is now achievable without any 3D assets, by cleverly transforming 2D images into effective training data.

Frontier LLMs struggle to replicate expert consensus in complex, longitudinal cancer treatment discussions, scoring below expert equivalence on a new real-world benchmark.

Enterprise AI agents can now generate valid, high-fidelity tabular data for training and evaluation without needing access to sensitive schemas or real systems, bypassing critical industry bottlenecks.

Synthetic-to-real pretraining bridges the chronic scarcity of longitudinal clinical imaging, enabling transformers to reliably forecast up to 15 years of spinal disease progression.

Purely synthetic training data outperforms native real-world supervision on satellite stereo benchmarks, showing that pristine geometric labels and zero temporal noise beat noisy real-world data collection.

PhyVisGen, a physically and visually high-fidelity framework for scalable robotic manipulation data generation, introduces an arm-gripper coupling method based on the Incremental Potential Contact (IPC), enabling high-fidelity soft contact throughout complete manipulation trajectories.

A conceptual framework for participatory speech data curation by, for, and with marginalized communities drawing on insights from co-design and knowledge sharing is developed, comprising overlapping and two-way processes of defining a community, project formulation, modes of participation, and personal autonomy.

This paper introduces time-varying exponentiated convolution methods that transform both Gaussian noise and impulse responses into reverberation and modified spectral-decay fields respectively and derives two recursive and fast convolution algorithms that extend into the spherical harmonic domain.

This work proposes UniK, a common platform for both digital and physical AI classes, covering the full knowledge lifecycle across modalities from rich text and video to molecular data and sensor telemetry, built on Polymath Retrieval with no task-specific fine-tuning.

EdgeGen is a synthetic task generation framework that extracts compliance rules from an agent's specification and uses them to generate database-grounded edge-case tasks designed to violate these rules, and when combined with existing synthetic data generation techniques, enables agent improvement through finetuning and harness optimization.

Protein annotation doesn't have to be trapped in rigid ontologies: a 4-bit 3B LLM can generate curator-grade functional descriptions directly from raw amino acid sequences.