Cerca nella ricerca
Trova paper e lezioni per titolo, tema o parola chiave.
62 risultati
Antenna agreement helps recover breathing under motion interference
A spatial adaptation method improves breathing-waveform recovery under simulated motion interference and lowers failures after real-data calibration.
Motion fakes leave clues MoDA uses to detect and trace
MoDA combines visual meaning and fine detail to detect motion deepfakes, including clips from two unseen video platforms.
A lattice method predicts better decoding schedules before generation
A lattice method estimates lost position dependence to rank decoding schedules before generation, with tested image evidence.
A tutor's higher fit score came with lower educator ratings
Educators rated corrected proxy plans worse than original tutor plans despite higher instructional-fit scores.
TokenRouter helps models hand off requests without restarting work
TokenRouter coordinates model handoffs and raised output-token rates over existing systems in tested workloads.
Human motion helps humanoid robots follow instructions
VioLA uses shared human and robot motion targets to train humanoid control, with strong results in one laboratory evaluation.
Reusing a student's first answers can improve distillation
Semi-on-policy distillation reuses a student's starting answers and outperformed standard distillation on 14 of 17 tested pairs.
Reaching the Goal, Responding in Time, and Staying Safe
What Else Changes When You Fix a Model?
Rewarding diverse robot successes improved tested shift averages
DRIVE rewards diverse successful robot behavior; simulation and two physical tasks show higher average success under tested shifts.
A human-trained user simulator improves agent training
MIMESIS models varied human behavior; agents trained with it score better than GPT-5.5-trained agents across nine simulated users.
Branching diffusion cuts the work of long text generation
Branching Latent Diffusion cuts computation for long text generation, while quality and single-sequence speed remain bounded.
SoftmaxLift reduces key retrieval for long-context attention
SoftmaxLift augments keys and queries to estimate attention with one exact search index and O(1/ε²) retrieved keys.
A rule for splitting images can preserve object-finding recall
A theory and overlapping-view rule explain when splitting large images can preserve object-finding recall under stated conditions.
An optimizer change slows forgetting during adaptation
A training-time optimizer adjustment reduced old-domain validation-loss increases in seven stable Korean adaptation settings.
ChatGPT help appeared in few research tasks, concentrated among workers
Linked ChatGPT and Prolific records showed assistance in about 1% of eligible task windows among weekly ChatGPT users.
A graph result powers faster structured language decoding
A GPU decoding engine recomputes format masks and speeds six JSON extractions on one tested laptop.
A direct formula makes private-training noise calibration less conservative
A direct noise-calibration method bounds membership tests in random-allocation DP-SGD and computes an exact reference.
COPC corrects stale weights and learning signals in language-model training
COPC corrects stale learning signals, with higher math and search accuracy than AReaL in two reported settings.
CanonQ trains language models to use one shared low-bit codec
CanonQ improves measured accuracy and text prediction for a tested language model under joint low-bit compression.
A reviewed MRI workflow finds duplicate identities across datasets
HAPPEN combines exact scan matching and human review, finding cross-dataset duplicates in an audited MRI repository.
Juno connects robot predictions to motion and action learning
Juno links motion-aware robot predictions to action learning, with gains in tested simulation tasks and physical shifts.
Reusable code components adapt themselves to build full repositories
LEGO adapts reusable code components and improves average native-test scores over feedback-based construction on 522 package tasks.
LayerRoPE turns layer depth into a signal for Transformer blocks
LayerRoPE generates Transformer normalization gains from shared, depth-adjusted values and had the lowest C4 validation loss at tested scales.