Indian Institute of Technology Delhi

QRAFT splits fact-checking article writing into evidence gathering, drafting and a simulated editorial review run by multiple LLM agents, beating prior text-generation methods on automatic metrics across 1,227 examples while still scoring well below expert-written articles with professional fact-checkers.

QRAFT splits fact-checking article writing into evidence gathering, drafting and a simulated editorial review run by multiple LLM agents, beating prior text-generation methods on automatic metrics across 1,227 examples while still scoring well below expert-written articles with professional fact-checkers.

LinSlot introduces a probabilistic generative model linking images, slots (objects) and blocks (attributes), imposes the linear representation hypothesis in both object and attribute spaces, and infers attributes from slots via block attention; on CLEVR-Easy, CLEVR-Hard and CLEVR-Tex it improves DCI and FG-ARI over slot attention, SLATE+ and SysBinder baselines, while providing evidence for global delta directions in slot space.

LinSlot introduces a probabilistic generative model linking images, slots (objects) and blocks (attributes), imposes the linear representation hypothesis in both object and attribute spaces, and infers attributes from slots via block attention; on CLEVR-Easy, CLEVR-Hard and CLEVR-Tex it improves DCI and FG-ARI over slot attention, SLATE+ and SysBinder baselines, while providing evidence for global delta directions in slot space.

For Muon under momentum accumulation and a tuned finite-step Newton–Schulz update, the work proves that full-batch training of a sufficiently wide two-layer ReLU network reaches any target empirical squared loss ε with high probability when the constant learning rate scales as (1-μ)√ε, with a hitting-time bound of O((1-μ)^-1ε^-1/2) and a sufficient width independent of both target accuracy and momentum.

For Muon under momentum accumulation and a tuned finite-step Newton–Schulz update, the work proves that full-batch training of a sufficiently wide two-layer ReLU network reaches any target empirical squared loss ε with high probability when the constant learning rate scales as (1-μ)√ε, with a hitting-time bound of O((1-μ)^-1ε^-1/2) and a sufficient width independent of both target accuracy and momentum.