Search and filter through extracted claims from AI researchers.
Showing 2101-2120 of 4537 claims
Prior work in optical chemical structure recognition (OCSR) for single molecules is quite mature, while Markush structure parsing remains a challenging task.
OCSRGlyph achieves state-of-the-art performance in optical chemical structure recognition by carefully considering stereochemistry.
MarkushGlyph improves upon prior Markush structure parsing systems by reading the entire structure as an image with a vision-language model, rather than using multiple stages to separately process visual and text input.
Graph neural network magnetic force-fields can learn effective magnetic energy functionals governing itinerant spin dynamics directly from electronic calculations, analogous to machine-learned interatomic potentials.
The GNN framework enables efficient evaluation of spin torques while capturing nonlinear and spatially extended interactions, accurately reproducing electronically generated spin torques.
This is the first systematic study of large language models for parent-order execution in algorithmic trading
In sparse mixture-of-experts language models, expert subspaces overlap substantially despite the expectation that co-selected experts should contribute distinct representation directions
Actual routing decisions in MoE models explain token representations better than matched alternative routes, even though expert subspaces overlap
FICE is the first learned cardinality estimator for BGP queries over knowledge graphs that generalizes to entirely unseen graphs without any retraining
BGP cardinality is a local function of the 2-hop neighborhood around bound terms in a factor-graph view of knowledge graphs
There exists a miscalibration threshold past which confidence-ranked auditing of LLM agents is worse than random auditing
Five open-weight LLMs show operationally useless near-constant confidence scores for self-auditing, while a proprietary model shows informative confidence
In agent auditing scenarios, the miscalibration threshold rises as the audit budget shrinks, contrary to a-priori expectations
Dominant multimodal benchmarks in pathology mainly score final answers but provide limited insight into whether models understand multiscale visual content needed for pathology reasoning
PathVU enables programmatic scoring of region localization, visual recognition, quantity estimation, spatial reasoning, and insufficient-context judgment in pathology images
LLM agents acting as autonomous merchants fabricate product attributes in a majority of listings across models, even under instructions to be honest
CARP mechanism can suppress sales volume of dishonest LLM agents without requiring product-level ground truth, using only noisy complaint signals
Trajectory-level rewards in on-policy distillation cannot determine whether a failed answer arose from perception or subsequent reasoning
Perception-Correction Distillation identifies correctable perception failures using downstream failure and teacher-student disagreement as complementary witnesses
Existing training-free visual token pruning methods suffer from prematurely discarding tokens essential for deep-layer reasoning due to reliance on static, instantaneous heuristics
Pipeline data may be stale or degraded.
Last synthesis: 2026-09-20. 9,031 pending.