eamag/beyond_benchmarks · Datasets at Hugging Face

we then inform simulations of a theoretical dynamical framework with several different

models

opinion dynamics

we then inform simulations of a theoretical dynamical framework with several different models for how populations ideology evolves over time including a model which reproduces current macro-scale ideological distributions given the empirical micro-scale data gathered

the model constrains the gas density placing the

wa

dense gas

the model constrains the gas density placing the wa cloud at 0

as photonic systems progress toward enhanced miniaturization dynamic reconfigurability and improved energy efficiency a central challenge endures the accurate and independent control of optical losses and resonant properties on

scalable

photonic circuits

as photonic systems progress toward enhanced miniaturization dynamic reconfigurability and improved energy efficiency a central challenge endures the accurate and independent control of optical losses and resonant properties on scalable cmos-compatible platforms

to address the non-convex and temporally coupled optimization problem we propose a mixture-of-experts-augmented soft actor-critic moe-sac

algorithm

reinforcement learning

to address the non-convex and temporally coupled optimization problem we propose a mixture-of-experts-augmented soft actor-critic moe-sac algorithm that employs a sparse top-k gated mixture-of-shallow-experts architecture to represent multimodal policy distributions arising from the conflicting optimization objectives

our results lay a fundamental understanding of how pre-trained llms manipulate numbers and outline the potential of more accurate probing

techniques

models llms

our results lay a fundamental understanding of how pre-trained llms manipulate numbers and outline the potential of more accurate probing techniques in addressed refinements of llms architectures

the current advancements in beam shaping techniques their impact on the nanoparticle characteristics and their broader implications for scaling pulsed

laser

pulsed laser

the current advancements in beam shaping techniques their impact on the nanoparticle characteristics and their broader implications for scaling pulsed laser ablation in liquids to meet industrial demands are highlighted offering a comprehensive perspective on the future of this dynamic field

we present a suite of 100 cosmologically motivated controlled n-body simulations designed to advance the understanding of the

role

dwarf galaxies

we present a suite of 100 cosmologically motivated controlled n-body simulations designed to advance the understanding of the role of purely gravitational dynamics in the early formation of low-mass galaxy groups 1-5 x 10 13 m_sun

the index is based on the principles of the 15-minute

city

large cities

the index is based on the principles of the 15-minute city paradigm moreno et al

but this method often fails to converge due to the system

s

zeroth-order methods

but this method often fails to converge due to the system s high dimensionality and nonlinearity

to overcome the potential non-smoothness of the hyper-objective and the computational challenges associated with the

hessian

accelerated gradient

to overcome the potential non-smoothness of the hyper-objective and the computational challenges associated with the hessian matrix we utilize penalty and augmented lagrangian methods to reformulate the original problem as a single-level one

i develop a nonparametric framework for identifying spatial

boundaries

effect boundaries

i develop a nonparametric framework for identifying spatial boundaries of treatment effects without imposing parametric functional form restrictions

we introduce the textbf boundless large model

blm

multi-goal visual

we introduce the textbf boundless large model blm _1 a multimodal spatial foundation model that preserves instruction following and reasoning incorporates embodied knowledge and supports robust cross-embodiment control

the instrument was built explicitly for research in human-ai interaction to measure

trust

trustworthy ai

the instrument was built explicitly for research in human-ai interaction to measure trust attitudes towards ai systems from layperson non-expert perspective

large language models llms have significantly advanced generative applications in

natural

natural language processing

large language models llms have significantly advanced generative applications in natural language processing nlp

this approach overcomes the ambiguity of qualitative inspection near regime boundaries particularly in large

systems

complex systems

this approach overcomes the ambiguity of qualitative inspection near regime boundaries particularly in large systems and provides a compact extensible framework for identifying and comparing emergent behaviors in complex systems

three-state coevolutionary game dynamics with

environmental

ecological interactions

three-state coevolutionary game dynamics with environmental feedback

online randomness extraction simulating barely random algorithms in the

random

online algorithm

online randomness extraction simulating barely random algorithms in the random order arrival model

self-localization based on a camera often uses a convolutional neural

network

computer vision

self-localization based on a camera often uses a convolutional neural network cnn that can extract local features that are calculated by nearby pixels

reinforcement learning rl fine-tuning of large

language

large language models llms

reinforcement learning rl fine-tuning of large language models llms often suffers from instability due to the numerical mismatch between the training and inference policies

a common approach is to train a machine learning model to predict counterfactual outcomes and then select the

policy

policy learning

a common approach is to train a machine learning model to predict counterfactual outcomes and then select the policy that optimizes the predicted objective value

speakers can often differentiate these loanwords from native vocabulary particularly in bilingual communities where a

dominant

large language

speakers can often differentiate these loanwords from native vocabulary particularly in bilingual communities where a dominant language continuously imposes lexical items on a minority language

we refer to this riesz representer estimation as generalized

riesz

riesz representer

we refer to this riesz representer estimation as generalized riesz regression

these quantum-inspired methods are expected to yield faster algorithms with applications ranging from astronomy and earth observation to microscopy and

classical

quantum dot

these quantum-inspired methods are expected to yield faster algorithms with applications ranging from astronomy and earth observation to microscopy and classical imaging more broadly

second we employ a debiasing procedure based on generalized jackknifing to enable inference with larger bandwidths while preserving convexity of the

objective

debiased machine learning

second we employ a debiasing procedure based on generalized jackknifing to enable inference with larger bandwidths while preserving convexity of the objective function

self-improvement has emerged as a mainstream paradigm for advancing the reasoning capabilities of

large

continual learning

self-improvement has emerged as a mainstream paradigm for advancing the reasoning capabilities of large vision-language models lvlms where models explore and learn from successful trajectories iteratively

large language models llms have demonstrated exceptional capabilities across

multiple

models llms

large language models llms have demonstrated exceptional capabilities across multiple domains by leveraging massive pre-training and curated fine-tuning data

vision-based end-to-end e2e driving has garnered significant interest in the research community due to its scalability and synergy with

multimodal

multimodal reasoning

vision-based end-to-end e2e driving has garnered significant interest in the research community due to its scalability and synergy with multimodal large language models mllms

building on this algorithm we next give a one-sided adaptive algorithm for this problem that does not need to be given the value of n and with high probability makes tilde o log n epsilon

samples

randomized algorithm

building on this algorithm we next give a one-sided adaptive algorithm for this problem that does not need to be given the value of n and with high probability makes tilde o log n epsilon samples and queries

allowing the order of quantum operations to exist in superposition is known to

open

quantum batteries

allowing the order of quantum operations to exist in superposition is known to open new routes for thermodynamic tasks

our results reveal temporal stability in aggregate

route

route choice

our results reveal temporal stability in aggregate route choice behavior across the entire urban region throughout 2023

we show how this composite optimization problem can be reduced to an optimization problem over the banach space component only up to a

linear

optimization problem

we show how this composite optimization problem can be reduced to an optimization problem over the banach space component only up to a linear problem

scale invariance and statistical significance in complex

weighted

scale-free networks

scale invariance and statistical significance in complex weighted networks

28 higher accuracy substantially outperforming frontier models such as claude 4 and openai o3 in pairwise

reward

reward density

28 higher accuracy substantially outperforming frontier models such as claude 4 and openai o3 in pairwise reward judgments

by distributing and recollecting the quantum state with an entanglement-distribution operation the scan rate scales as

n

quantum computing

by distributing and recollecting the quantum state with an entanglement-distribution operation the scan rate scales as n 2 m 1 while thermal excitation is the dominant background significantly outperforming classical single-cavity methods under matched conditions

while resting-state fmri studies have revealed large-scale network correlates of creative potential electroencephalography

eeg

brain activity

while resting-state fmri studies have revealed large-scale network correlates of creative potential electroencephalography eeg offers a temporally precise and scalable approach to capture the fast oscillatory dynamics that underlie spontaneous neural organization

we investigate a quantum heat engine where energy exchanges are driven by generalized measurements and the sequence of these

operations

quantum networks

we investigate a quantum heat engine where energy exchanges are driven by generalized measurements and the sequence of these operations is coherently controlled in a superposition of causal orders

from 1975 to 2025 urban populations have become increasingly concentrated in

large

urban systems

from 1975 to 2025 urban populations have become increasingly concentrated in large cities

neuronmm high-performance matrix multiplication for llm

inference

llm inference

neuronmm high-performance matrix multiplication for llm inference on aws trainium

on the limitation of evaluating machine unlearning using only a single

training

machine learning

on the limitation of evaluating machine unlearning using only a single training seed

reinforcement learning rl can elicit strong reasoning in large language models llms yet most open efforts focus on

math

reasoning curriculum

reinforcement learning rl can elicit strong reasoning in large language models llms yet most open efforts focus on math and code

together these results suggest that higher immersion amplifies both the benefits and costs of

sensorimotor

sensorimotor disruptions

together these results suggest that higher immersion amplifies both the benefits and costs of sensorimotor coherence

a closed-form approximation-free control law is derived to

ensure

control law

a closed-form approximation-free control law is derived to ensure that each agent remains within its evolving stt thereby avoiding dynamic obstacles while also preventing inter-agent collisions in a socially aware manner and reaching the target within a prescribed time

combining moving mass actuators and manoeuvring models for

underwater

moving mass

combining moving mass actuators and manoeuvring models for underwater vehicles a lagrangian approach

our method extends prior work by decoupling the upper and lower bounds enabling more flexible and

tighter

computationally efficient

our method extends prior work by decoupling the upper and lower bounds enabling more flexible and tighter approximations

an optimal solution can be found with a greedy algorithm steel systematic

biology

phylogenetic tree

an optimal solution can be found with a greedy algorithm steel systematic biology 2005 pardi and goldman plos genetics 2005

results showed that revision behaviour differed across

feedback

directive metacognitive

results showed that revision behaviour differed across feedback conditions with hybrid prompting the most revisions compared to directive and metacognitive

assessing scenario coverage is crucial for evaluating the robustness of autonomous agents yet existing methods rely on expensive human annotations or computationally intensive large

vision-language

vision-language-action vla

assessing scenario coverage is crucial for evaluating the robustness of autonomous agents yet existing methods rely on expensive human annotations or computationally intensive large vision-language models lvlms

we therefore extend it by developing a local representation that applies to joint

probabilities

categorical outcomes

we therefore extend it by developing a local representation that applies to joint probabilities thereby eliminating the need to impose an artificial ordering on categories

a freeable matrix characterization of bipartite

graphs

bipartite graphs

a freeable matrix characterization of bipartite graphs of ferrers dimension three

evaluated on a comprehensive dataset from the canadian prairies cypress demonstrates superior performance over existing deep learning-based

yield

yield prediction

evaluated on a comprehensive dataset from the canadian prairies cypress demonstrates superior performance over existing deep learning-based yield prediction models highlighting the effectiveness of fine-tuning foundation models for specialized agricultural applications

these limitations are particularly apparent in real-life driving scenarios where state-of-the-art algorithms struggle to safely or reliably complete

overtaking

collision avoidance

these limitations are particularly apparent in real-life driving scenarios where state-of-the-art algorithms struggle to safely or reliably complete overtaking manoeuvres

meanwhile it provides a unified theoretical explanation for classic topological characteristics such as small-world networks and

scale-free

network structures

meanwhile it provides a unified theoretical explanation for classic topological characteristics such as small-world networks and scale-free networks

masiero lifting partial smoothing to solve hjb equations and stochastic control problems siam journal on

control

optimal control

masiero lifting partial smoothing to solve hjb equations and stochastic control problems siam journal on control and optimization 63 3 2025 pp

using a brain encoder as a digital twin offers a powerful data-driven framework for generating and testing hypotheses about visual selectivity in the human brain - hypotheses that can guide future

fmri

fmri data

using a brain encoder as a digital twin offers a powerful data-driven framework for generating and testing hypotheses about visual selectivity in the human brain - hypotheses that can guide future fmri experiments

where a bayesian treatment is usually associated with high-quality predictions and uncertainties the practical reality has been the opposite with unstable training poor

predictive

predictive performance

where a bayesian treatment is usually associated with high-quality predictions and uncertainties the practical reality has been the opposite with unstable training poor predictive power and subpar calibration

an n 1 -bit toffoli gate is mainly utilized to construct other quantum gates and operators such as fredkin

gates

-bit toffoli

an n 1 -bit toffoli gate is mainly utilized to construct other quantum gates and operators such as fredkin gates arithmetical adders and logical comparators where n geq 2

specifically we propose a thinking protocol where an organizer dynamically assigns sub-queries to workers merges

intermediate

thinking traces

specifically we propose a thinking protocol where an organizer dynamically assigns sub-queries to workers merges intermediate knowledge and produces coherent solutions

trainium an ai accelerator recently developed by amazon web services

aws

llm agents

trainium an ai accelerator recently developed by amazon web services aws provides an attractive option for llm training and inference through its heterogeneous architecture

we investigate the thermal evolution of 3i atlas the third macroscopic

interstellar

circumgalactic medium

we investigate the thermal evolution of 3i atlas the third macroscopic interstellar object discovered on 2025 july 1

5 is pre-trained end-to-end with a unified next-token prediction objective on a corpus of

vision-language

vision-language models vlms

5 is pre-trained end-to-end with a unified next-token prediction objective on a corpus of vision-language interleaved data containing over 10 trillion tokens primarily derived from sequential frames and transcripts of internet videos

through extensive experiments and human evaluations we show that our approach not only enhances alignment between explanation and prediction but also empowers mllms to deliver emotionally coherent trustworthy

interactions

human-ai interaction

through extensive experiments and human evaluations we show that our approach not only enhances alignment between explanation and prediction but also empowers mllms to deliver emotionally coherent trustworthy interactions marking a key step toward truly human-like hci systems

this study experimentally tested whether short-term exposure to narrow ai tools enhances core

cognitive

ai systems

this study experimentally tested whether short-term exposure to narrow ai tools enhances core cognitive abilities or simply optimizes task performance

through iterative debates the agents progressively refine their proposals producing increasingly effective

robot

learning agents

through iterative debates the agents progressively refine their proposals producing increasingly effective robot designs

gradient flow sampler-based distributionally

robust

gradient flow

gradient flow sampler-based distributionally robust optimization

we study policy learning with abstention where a

policy

policy learning

we study policy learning with abstention where a policy may defer to a safe default or an expert

by converting each dataset into interpretable metadata we prompt an

llm

models llms

by converting each dataset into interpretable metadata we prompt an llm to recommend both model families and hyperparameters

the growing integration of artificial intelligence ai into human cognition raises a fundamental

question

ai use

the growing integration of artificial intelligence ai into human cognition raises a fundamental question does ai merely improve efficiency or does it alter how we think

hybrid consistency policy decoupling multi-modal diversity and real-time efficiency in

robotic

robotic systems

hybrid consistency policy decoupling multi-modal diversity and real-time efficiency in robotic manipulation

using a k-spine as a central guide we introduce an o klog dist exponential search algorithm on a

tree

tree edit distance

using a k-spine as a central guide we introduce an o klog dist exponential search algorithm on a tree by searching mainly along the spine to narrow down the target s vicinity and then recursively handling the smaller components

preference learning from pairwise feedback is a widely adopted framework in applications such as reinforcement learning with human

feedback

preference learning

preference learning from pairwise feedback is a widely adopted framework in applications such as reinforcement learning with human feedback and recommendations

here we present phyloformer 2 the first likelihood-free

inference

phylogenetic tree

here we present phyloformer 2 the first likelihood-free inference method for posterior distributions over phylogenies

our model interpolates between classical models of fitness

waves

traveling waves

our model interpolates between classical models of fitness waves and exhibits a novel phase transition in the propagation of the wave

why do human populations remain vulnerable to

collapse

large population

why do human populations remain vulnerable to collapse even when they are large

our results provide the most direct empirical evidence to date for the connection between galaxy spins and the cosmic

tidal

tidal field

our results provide the most direct empirical evidence to date for the connection between galaxy spins and the cosmic tidal field

firstly this paper presents an emergency-aware reasoning framework which dynamically adjusts

reasoning

reasoning curriculum

firstly this paper presents an emergency-aware reasoning framework which dynamically adjusts reasoning depth based on the emergency scenario and is equipped with a novel reviewer-based emergency rag rerag to distill specific knowledge and guidance from historical cases enhancing the reliability and rationality of agents emergency decisions

our spg-cdenet consists of two key components a spatial prior network and a cross dual

encoder

encoder network

our spg-cdenet consists of two key components a spatial prior network and a cross dual encoder network

using the next generation matrix method we derive an analytical expression for the

basic

basic reproduction number

using the next generation matrix method we derive an analytical expression for the basic reproduction number mathcal r_0

herd immunity requires vaccination of approximately seventy percent of the population but this increases to circa eighty

percent

adaptive immune

herd immunity requires vaccination of approximately seventy percent of the population but this increases to circa eighty percent for the more transmissible alpha-variant

vision-language models vlms exhibit uneven performance across

languages

language models

vision-language models vlms exhibit uneven performance across languages a problem that is often exacerbated when the model size is reduced

63 times speedup offering a practical solution for

efficient

models llms

63 times speedup offering a practical solution for efficient llm deployment

we design meta-algorithms that reduce the problem to the

unweighted

randomized algorithm

we design meta-algorithms that reduce the problem to the unweighted approximate maximum cardinality matching mcm problem

we then apply these results in a filtering scenario to analyze the optimal

transport

optimal transport

we then apply these results in a filtering scenario to analyze the optimal transport filtering algorithm of al-jarrah et al

these findings raise fundamental questions about when and how well-known structural components of galaxy morphology such as

bulges

massive stars

these findings raise fundamental questions about when and how well-known structural components of galaxy morphology such as bulges and disks first emerged

tree embedding has been a fundamental method in algorithm design with

wide

spanning trees

tree embedding has been a fundamental method in algorithm design with wide applications

this study investigates the performance of model predictive control mpc and rule-based

control

predictive control

this study investigates the performance of model predictive control mpc and rule-based control rbc under 15 30 60 minute averaging commonly used in research when net billing and battery degradation are considered

the vacuum-gap optical cavity operates in air

achieving

optical interference

the vacuum-gap optical cavity operates in air achieving a quality factor of approximately 2

galaxy mergers trigger starburst activity and galactic outflows that enrich the

circumgalactic

circumgalactic medium

galaxy mergers trigger starburst activity and galactic outflows that enrich the circumgalactic medium profoundly impacting galaxy evolution

however these evaluations rely on llms proxy

llms

proxy llms

however these evaluations rely on llms proxy llms to gauge compliance with privacy norms overlooking real users perceptions

third traditional difference-in-differences methods that ignore

spatial

treatment effect boundaries

third traditional difference-in-differences methods that ignore spatial and network structure exhibit 61 percent bias in estimated treatment effects

the device consists of an array of unit cells each containing a loop with multiple jjs that amplifies weak quantum signals near the quantum

noise

fault-tolerant quantum

the device consists of an array of unit cells each containing a loop with multiple jjs that amplifies weak quantum signals near the quantum noise limit

we evaluate our approach on real-world datasets and

demonstrate

real-world datasets

we evaluate our approach on real-world datasets and demonstrate that integrating interventional constraints not only improves model accuracy and ensures consistency with established findings making models more explainable but also facilitates the discovery of new causal relationships that would otherwise be costly to identify

during massive star formation dense gas undergoes

chemical

star formation rates

during massive star formation dense gas undergoes chemical evolution producing both simple and complex organic molecules coms characteristic of hot molecular cores

when treatment effects are heterogeneous however such specifications generally fail to recover this

average

average treatment

when treatment effects are heterogeneous however such specifications generally fail to recover this average effect

instead of sharing a global low-resolution space each part in our method - even small ones - is

generated

image generation

instead of sharing a global low-resolution space each part in our method - even small ones - is generated at full resolution enabling the synthesis of intricate details

we present structlayoutformer a novel transformer-based approach for conditional structured

layout

layout generation

we present structlayoutformer a novel transformer-based approach for conditional structured layout generation

the framework enables natural gradient descent optimization

aligned

gradient flow

the framework enables natural gradient descent optimization aligned with the learned manifold geometry and offers unprecedented interpretability by endowing internal representations with clear geometric meaning

this framework can be applied to policy evaluation using the panel data approach pda where we further establish inference for the

average

treatment effect boundaries

this framework can be applied to policy evaluation using the panel data approach pda where we further establish inference for the average treatment effect

the coordination game payoff structure captures the insight that mutualistic

strategies

evolutionary game

the coordination game payoff structure captures the insight that mutualistic strategies lead to robust advantages only after such biological markets reach a certain scale

the problem is known to be substantially harder than the k -server problem and prior to this work even for k 3 taxis it has been unknown whether a finite competitive

ratio

competitive ratio

the problem is known to be substantially harder than the k -server problem and prior to this work even for k 3 taxis it has been unknown whether a finite competitive ratio is achievable on general metric spaces

our key idea is to decouple visual and linguistic adaptation by introducing two lightweight modules a domain classifier to identify the input image type and a dual adapter mechanism comprising a prompt adapter for

language

multimodal reasoning

our key idea is to decouple visual and linguistic adaptation by introducing two lightweight modules a domain classifier to identify the input image type and a dual adapter mechanism comprising a prompt adapter for language modulation and a visual adapter for vision feature adjustment