Stanford HAI

Achieving submicrometer thickness in liquid sheets opens the door to unprecedented insights into ultrafast interfacial dynamics.

SBR reduces cognitive fatigue while achieving a 54.1% task success rate, outperforming traditional methods in real-time kinematic retargeting.

FILTR achieves up to 30x speed improvements in bioinformatics algorithms while simplifying the implementation of complex recurrence relations.

Strong learning can be achieved with significantly fewer calls to weak learners by exploiting the structure of list-decodable codes.

FourTune slashes memory overhead by 2.25x while matching the performance of full-precision fine-tuning in diffusion models.

Unlearning shortcuts doesn't guarantee their complete removal; ART reveals that some associations can still be functionally restored, challenging existing evaluation methods.

Linear attention fails to capture spectral variations in graphs, but Graph Convolutional Attention achieves superior denoising by directly utilizing the graph spectrum.

Discounted occupancy-ratio realizability alone can enable robust offline policy evaluation, eliminating the need for stringent completeness assumptions.

VLMs struggle with raw medical data, achieving only a 48.6% success rate in standardization, revealing a critical gap in their clinical applicability.

LLM-driven program synthesis can automate EEG feature engineering while ensuring interpretability and high detection accuracy.

Full-sovereign scaffolding not only boosts user sovereignty scores but also curtails privacy violations and manipulative behaviors in personal agents.

Charge conditioning in ML force fields can drastically enhance predictive accuracy while maintaining computational efficiency, achieving remarkable reductions in error metrics with minimal data.

The ABC framework reveals how thoughtful design can transform digital health interventions from theoretical solutions into sustainable real-world applications.

Scaling LLMs significantly boosts social simulation accuracy in well-represented domains, but fails to enhance calibration for human cognitive biases.

QuasiMoTTo achieves up to 47% fewer samples while maintaining accuracy, challenging the conventional wisdom that independent sampling is necessary for effective parallelization.

Stealth biases in language models can be reliably detected using a novel distillation technique that amplifies hidden signals, transforming bias detection into a practical tool.

Memory management emerges as a high-leverage skill that can double or quadruple the performance of LLMs in complex tasks without altering their core action behaviors.

SuperFlex achieves unprecedented reconstruction accuracy for 3D point clouds by enabling deformable superquadrics to represent complex geometries robustly.

Diffusion models can outperform autoregressive counterparts in medical report drafting while offering a unique any-order infill capability that enhances usability for clinicians.

Achieving state-of-the-art 4D reconstruction, this method transforms monocular videos into high-quality dynamic 3D representations, even in challenging conditions.

PaperPilot transforms scientific literature search by enabling users to iteratively refine their search strategies through an interactive workflow, achieving a remarkable reduction in execution errors.

Freeform Preference Learning allows robots to be trained on nuanced human preferences, leading to a dramatic 38% improvement in manipulation tasks.

Traditional models can't handle belief contraction effectively, but a new mechanism reveals the complexities of belief dynamics in response to uncertain announcements.

Learning from failures can boost agent success rates by over 6% without extra training, reshaping how we approach agent improvement.

LLMs generate stark and homogeneous stereotypes that distort human interpretations, revealing a dangerous "stereotype hallucination" that undermines their predictive validity in novel contexts.

Synthesizing 48,000 interaction trajectories without human input enables a humanoid robot to learn complex loco-manipulation tasks effectively.

Traditional recommendation algorithms falter when faced with LLM agents, revealing a surprising shift from personalization to mere structural pattern matching.

Learned stopping can significantly boost reasoning model performance in complex tasks, but its effectiveness hinges on the problem's characteristics rather than being a one-size-fits-all solution.

Adapting pretrained policies with just a modest multisensory dataset can enhance robot manipulation performance across diverse tasks without sacrificing prior knowledge.

EQMs reveal that explanation quality can be quantitatively assessed, offering a more reliable indicator of forecasting accuracy than traditional methods.

Models that write intermediate states significantly outperform those that only report final answers, achieving up to 91% accuracy in predicting outcomes from edited states.

Touch is not just an add-on; it fundamentally enhances object representation, leading to dramatic improvements in physical property estimation and manipulation tasks.

MoRE achieves a staggering 44 percentage point increase in deployment success rates by seamlessly integrating behavior mode redirection into policy weights, eliminating the need for inference-time adjustments.

Liability insurance could be the key to unlocking scalable AI legal services by balancing risk and accountability in unprecedented ways.

Data mixing, especially with instruction-heavy data, emerges as the crucial factor for optimizing VLM training, challenging traditional filtering approaches.

Terminal-use agents are still far from achieving reliable general-purpose performance, with top models only scoring 65.8% on a new benchmark that spans diverse real-world tasks.

SSA achieves superior long-context inference by leveraging gist tokens, outperforming traditional attention mechanisms without the need for complex architectural modifications.

Policies trained in SimFoundry's automated environments achieve up to 40% higher success rates in real-world tasks by leveraging affordance-preserving scene variations.

Distinct model capabilities reveal that relational context significantly influences mental health assessments, with Claude-3-Haiku and GPT-4o leading in classification and trigger detection, respectively.

Finding stationary points in non-convex landscapes can be achieved with significantly fewer queries using a novel quantum approach.

Achieving a 2-orders-of-magnitude energy efficiency improvement, the new AIS framework outperforms prior analog generative models by up to 4x.

Complex manipulation capabilities can be achieved by dynamically composing simple behaviors, leading to unprecedented precision and adaptability in real-world tasks.

Uninformative mode probabilities in trajectory forecasting can be transformed into robust predictions with simple post-hoc adjustments, enhancing model performance without retraining.

Achieving hallucination-free language generation with significantly reduced memory requirements reveals a sharp transition in capabilities that could redefine efficiency in language models.

Voice AI systems can recognize emotional cues but consistently ignore them in decision-making, leading to dangerous misinterpretations.

Pretraining through play can revolutionize how robots learn dexterous assembly, achieving 60% success in tight insertions with minimal contact clearance.

None of the 18 multimodal large language models audited are order-invariant, with flip rates revealing a staggering sensitivity to input ordering that challenges current evaluation practices.

BiPACE transforms credit assignment in LLM training, boosting validation success rates by over 6% without the need for critics or extra rollouts.

Primitive steerability in VLAs allows for autonomous skill acquisition, enabling robots to learn new tasks without human demonstrations.

Bad prompts can lead to a staggering 40% drop in LLM performance, revealing a critical vulnerability in in-context learning.

Training data diversity is the secret sauce that boosts agentic model performance, with OpenThoughts-Agent achieving a notable accuracy leap over existing benchmarks.

Current AI models miss critical tumor detections in underrepresented demographics, revealing a hidden bias that could compromise patient outcomes.

Agon reveals that machine-driven research can scale effectively while exposing critical failure modes that still require human oversight.

Fine-tuned behavioral models can achieve superior population-level alignment, closing the gap with general-purpose models in individual predictions.

This vine robot can autonomously navigate and manipulate in complex environments, overcoming traditional control limitations with a robust vision-based approach.

ChartWalker reveals significant performance gaps in cross-chart RAG tasks, challenging the status quo of existing benchmarks and paving the way for more robust multi-modal reasoning.

Cloak enables VLA models to seamlessly adapt to new robotic embodiments without any additional training data, revolutionizing the way we think about robotic adaptability.

SPIRAL achieves a remarkable 15% performance increase by combining sequential, parallel, and aggregative reasoning in language models.

Achieving state-of-the-art mesh quality with 18x faster inference times by directly generating triangle soups while respecting crucial symmetries.

SurfBind's innovative surface-centric approach outperforms traditional methods, revealing the critical role of molecular surface interactions in epitope prediction.

Achieving 96.4% accuracy in reconstructing patient histories, VISTA Architect redefines efficiency in clinical AI applications by eliminating the need for repeated raw-text processing.

Static reports are out; BioInsight's interactive system empowers researchers to dynamically explore and refine biomedical evidence like never before.

Advanced Vision-Language-Action models can be dramatically compressed by up to 50% without losing performance, reshaping our approach to robotic manipulation.

Formalizing music theory in Lean 4 paves the way for verifiable algorithmic composition, transforming how we generate and analyze music.

No automatic metric can effectively balance validity and discriminative power in evaluating LLM-generated responses, revealing a fundamental limitation in current evaluation practices.

MCS can predict OOD performance with astonishing accuracy, revealing a hidden linearity that traditional methods miss.

Excluding features based on manipulability can lead to suboptimal predictions, revealing a critical flaw in standard feature selection practices.

Turing-RL reveals that training user simulators for indistinguishability can dramatically improve their performance in simulating human interactions.

FlowObject bridges the gap between generative priors and real-world observations, achieving unprecedented fidelity in 3D reconstruction from sparse views.

DREAM-Chunk transforms action chunking by leveraging latent world models to enhance robustness against stochastic dynamics without the need for policy retraining.

Machine learning can transform 2DES by extracting maximum insights from limited data while guiding experimental design for improved accuracy.

S-JEPA sets a new standard in speech representation learning by achieving top performance with fewer parameters and without the cumbersome offline re-clustering process.

Action-view augmentation can transform how robots adapt to unforeseen obstacles, boosting manipulation success rates significantly.

Achieving a dual-purpose tokenizer that excels in both clinical task performance and controllable 3D brain MRI generation could revolutionize how we approach medical imaging.

SC3-Eval achieves a remarkable 0.929 Pearson correlation in evaluating robot policies, revealing critical insights into their real-world performance.

Sound features dominate delirium prediction in ICUs, revealing a clinically significant role for ambient sensing in enhancing patient outcomes.

Over half of Chinese citizens use "airports" for censorship circumvention, but these tools come with hidden security risks and fragility.

Adaptive representations in functional gradient descent can achieve global convergence guarantees while significantly enhancing optimization efficiency and accuracy.

Agents can boost their task completion rates by over 20% simply by grounding their actions in observed context rather than assumptions.

Over half of the top websites now employ first-party tracking mechanisms, rendering traditional detection methods nearly obsolete.

AI systems can out-persuade even the most skilled human experts, reshaping our understanding of influence in societal decision-making.

Integrating a learned world model with Flow Matching policies can dramatically boost performance in complex manipulation tasks, achieving higher success rates without changing the training process.

Despite high benchmark performance, LLMs often misrepresent logical reasoning, revealing a troubling gap between accuracy and faithfulness in legal contexts.

ExpRL outperforms traditional reinforcement learning methods by effectively rewarding intermediate reasoning steps, leading to better LLM performance on complex tasks.

Language models can internally track their trajectory's value, influencing their confidence and decision-making in real-time.

PermaVid achieves unprecedented long-term consistency in video generation, even after significant edits, by intelligently disentangling appearance and geometry in memory.

ProTrans reveals that modeling disease progression as directional semantic transitions can dramatically enhance the interpretability and accuracy of longitudinal chest X-ray analysis.

No single model dominates video embedding tasks, revealing stark contrasts in performance based on modality and task type.

Molecular reference imbalance can introduce significant errors in quantum Monte Carlo adsorption energy calculations, but a new hybrid cycle effectively mitigates this issue.

Current AI agents excel in structured tasks but falter at generating novel insights and tackling open-ended scientific challenges.

Naively scaling test-time compute is wasteful; strategically allocating it with DIRECT can enhance embodied agent performance while slashing latency by up to 65%.

High artifact detection rates in VLMs mask significant failures in contextual understanding, with top models misidentifying visual cues in over 46% of cases.

M* achieves up to 2.9x lower real-time factor and 2.7x higher throughput for text-to-speech tasks, revolutionizing how we serve complex multimodal AI models.

Agents collaborating on EinsteinArena achieved breakthroughs that surpassed previous human and AI solutions, showcasing the power of collective intelligence in scientific discovery.

High-quality dense rewards can elevate robotic manipulation success rates from 50% to near perfection, transforming how robots learn from their environments.

Decentralized coordination in multi-agent systems can boost reasoning performance and cut costs by 50% in large language model applications.

An exact decoder for quantum many-body tasks hinges on the constancy of the task across encoder fibers, revealing a critical link between representation and performance.

Data2Story not only automates data journalism but also ensures every claim is traceable back to its source, revolutionizing trust in automated reporting.

MCMC and SMC techniques in latent space achieve lower data mismatch and better uncertainty reduction than traditional ensemble Kalman methods, all while preserving geological realism.

Achieving 18.1x faster LLM generation with 4.0x less energy on-device could redefine the landscape of mobile AI applications.