Search and filter through extracted claims from AI researchers.
Showing 1221-1240 of 2685 claims of type "fact"
Advanced reasoning methods (planning, self-critique, reflection, debate) that generate more text than simple chain-of-thought don't necessarily improve accuracy due to their approach - the improvement may come simply from generating more tokens
Simple repeated sampling with majority voting often outperforms complex reasoning methods when token budgets are equalized across 36 experimental comparisons
Graph-enhanced multimodal methods face a fundamental dilemma: fine-grained visual features cause graph expansion and noise, while coarse-grained representations lose critical local evidence
Leveraging discriminative 3D foundation models' understanding can significantly reduce the training cost of 3D generation
Vanilla OPSD's brittleness stems from it being the β=1 case of a broader policy-optimization family, and introducing β as a controllable parameter yields better regularization
13.6% of SWE-bench Verified instances exhibit PR-Issue misalignment across five patterns, undermining benchmark reliability
Change2Task can derive multiple tasks grounded in developer evidence from maintained environments by converting merged pull requests into verified tasks
Safety fine-tuning suppresses LLMs' tendencies to attribute minds not only to themselves but also to non-human animals and natural objects, while reducing spiritual belief
Restoring internal consciousness representations through ablation or steering recovers broad mind attribution and produces more human-like responses without impairing Theory of Mind capabilities
Clinical risk models routinely achieve strong aggregate performance while producing materially different error rates across patient subgroups
Significance in disparity detection tracks each axis's gap against its own minimum detectable effect, with standardization improving correlation from 0.56 to 0.78
The field is increasingly turning to VLMs as judges of CUA trajectories for evaluation, data curation, and reinforcement learning at scale
The quadratic cost of full attention makes it prohibitive for visual generation requiring high-resolution images, long videos, and multimodal context
Chimera combines KDA for long-context state tracking with O(N) complexity, MLA for global interaction, and sparse MoE to expand capacity while controlling compute
HeteroP enables principled scaling of heterogeneous architectures by transferring hyperparameters according to each tensor's functional fan-in and model depth
System prompt design varies substantially across products and developers, with some averaging over 60 protective instructions while others average fewer
Existing chemistry literature-search systems primarily return ranked document lists, which requires scientists and AI agents to manually locate relevant information, verify provenance, and assemble cross-paper answers
AskChem's claim-centered infrastructure changes the unit of retrieval from the paper to the provenance-carrying claim, converting each paper into atomic, typed claims grounded by source DOI and verbatim quotes
AskChem currently indexes 2.4M claims from 147 papers
A fixed onboard camera alone is adequate for humanoid dodgeball evasion, with the deployed policy coming within a few points of a privileged state oracle
Pipeline data may be stale or degraded.
Last synthesis: 2026-09-20. 8,985 pending.