PulseAugur / Brief
EN
LIVE 16:36:01

Brief

last 24h
[2/2] 222 sources

Multi-source AI news clustered, deduplicated, and scored 0–100 across authority, cluster strength, headline signal, and time decay.

  1. TOOL · arXiv cs.LG English(EN) ·

    Knowledge-Informed Kernel State Reconstruction from Heterogeneous Partial Observations

    Researchers have developed MAAT, a framework designed to reconstruct the states of partially observed dynamical systems. This method operates within a reproducing kernel Hilbert space and integrates various observation types along with prior knowledge like non-negativity and conservation laws. MAAT has demonstrated significant improvements in trajectory and derivative reconstruction error across multiple scientific benchmarks and a real-world COVID-19 dataset. AI

    IMPACT Provides a new method for analyzing incomplete scientific data, potentially accelerating discovery in fields reliant on dynamical systems.

  2. RESEARCH · Hugging Face Daily Papers English(EN) · · [2 sources]

    MAAT: Multi-phase Adapter-Aware Targeted Unlearning

    Researchers have introduced MAAT, a novel three-phase framework for targeted machine unlearning that specifically addresses the difficulty of removing causal knowledge. Existing benchmarks are skewed, underrepresenting "Why" questions, which are crucial for evaluating causal and relational knowledge removal. MAAT operates on LoRA adapter weights and employs techniques like gradient-projected ascent and SVD pruning to achieve high forgetting while retaining other knowledge. The accompanying 5WBENCH benchmark, with balanced categories for Who, What, When, Where, and Why, quantifies these unlearning failures for the first time. AI

    IMPACT Introduces a new benchmark and method to improve the evaluation and execution of machine unlearning, particularly for causal knowledge.