prompt
stringlengths
41
511
target
stringlengths
1
25
keyword
stringclasses
697 values
full_sentence
stringlengths
48
1.25k
current tool-use large language models llms are trained on static datasets enabling them to interact with external
tools
large language
current tool-use large language models llms are trained on static datasets enabling them to interact with external tools and perform multi-step tool-integrated reasoning which produces tool-call trajectories
current data selection methods are largely heuristic-based lacking
theoretical
machine learning
current data selection methods are largely heuristic-based lacking theoretical guarantees and generalizability
in this work we study data-driven stabilization of linear time-invariant systems using
prior
data-driven stabilization
in this work we study data-driven stabilization of linear time-invariant systems using prior knowledge of system-theoretic properties specifically stabilizability and controllability
moreover we quantitatively interpret our components from a neuroscience perspective and analyze the associations between different visual
patterns
brain decoding
moreover we quantitatively interpret our components from a neuroscience perspective and analyze the associations between different visual patterns and brain functions
to characterize and measure these effects we probe vla s hidden representations and analyze attention maps further we design a set of targeted tasks and methods that contrast vla
models
vision-language-action vla
to characterize and measure these effects we probe vla s hidden representations and analyze attention maps further we design a set of targeted tasks and methods that contrast vla models with their counterpart vlms isolating changes in vl capabilities induced by action fine-tuning
however channel aging caused by user mobility and processing delays degrades the accuracy of
csi
channel state information csi
however channel aging caused by user mobility and processing delays degrades the accuracy of csi leading to suboptimal link adaptation and loss of performance
optical techniques for spatiotemporal control can produce
laser
nonlinear optical
optical techniques for spatiotemporal control can produce laser pulses with custom amplitude phase or polarization structure
cumulants moments and selection the connection between
evolution
evolutionary dynamics
cumulants moments and selection the connection between evolution and statistics
we apply this approach to three leading models i the general treatment model under unconfoundedness ii the negative control model and iii the long-term
causal
causal inference
we apply this approach to three leading models i the general treatment model under unconfoundedness ii the negative control model and iii the long-term causal inference model under unobserved confounding
large language model agent personality and response appropriateness evaluation by human linguistic experts llm-as-judge and natural
language
language models
large language model agent personality and response appropriateness evaluation by human linguistic experts llm-as-judge and natural language processing model
however realising electrical spin readout in
wide-bandgap
spin readout
however realising electrical spin readout in wide-bandgap materials with similar fidelity and bandwidth to optical approaches remains challenging
in this paper we train a neural network nn to solve the optimal power shutoff
line
line switching
in this paper we train a neural network nn to solve the optimal power shutoff line switching problem
dynamic context-aware scene reasoning using
vision-language
vision-language models
dynamic context-aware scene reasoning using vision-language alignment in zero-shot real-world scenarios
scale invariance and statistical significance in
complex
complex networks
scale invariance and statistical significance in complex weighted networks
while resting-state fmri studies have revealed large-scale network correlates of creative potential electroencephalography
eeg
fmri data
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
first we find that combining english and multilingual data does not necessarily degrade the in-language performance of either group provided that
languages
language agents
first we find that combining english and multilingual data does not necessarily degrade the in-language performance of either group provided that languages have a sufficient number of tokens included in the pretraining corpus
to enable our flow algorithms to work under vertex
capacities
maximum flow
to enable our flow algorithms to work under vertex capacities we also develop a close-to-linear time algorithm for computing length-constrained vertex expander decomposition
emph second we show that combining foundation models for time
series
time series classification
emph second we show that combining foundation models for time series forecasting with a reliability estimator can filter our unreliable predictions
in recent years the rapid advancements in artificial intelligence ai and machine learning ml have demonstrated remarkable potential in the modeling of nanoscale heat
conduction
heat conduction
in recent years the rapid advancements in artificial intelligence ai and machine learning ml have demonstrated remarkable potential in the modeling of nanoscale heat conduction and radiation
infected individuals in some epidemics can remain
asymptomatic
infectious individuals
infected individuals in some epidemics can remain asymptomatic while still carrying and transmitting the infection
we present simulation case studies to demonstrate the potential of the proposed approach in keeping the caev platoon operating safely without
collisions
collision avoidance
we present simulation case studies to demonstrate the potential of the proposed approach in keeping the caev platoon operating safely without collisions by curbing the effect of adversarial attacks
2023 shows that nn matching is an instance of density-ratio estimation with their new
density-ratio
density ratio
2023 shows that nn matching is an instance of density-ratio estimation with their new density-ratio estimator
rlmeval provides a new challenging benchmark designed to guide and accelerate progress in automated reasoning for
formal
mathematical reasoning
rlmeval provides a new challenging benchmark designed to guide and accelerate progress in automated reasoning for formal mathematics
here we develop a general theoretical framework that establishes quantitative relationships between the strength and timing of transient
dynamics
traffic dynamics
here we develop a general theoretical framework that establishes quantitative relationships between the strength and timing of transient dynamics to various inputs in heterogeneous networks
we formulate estimation of the nuisance parameters the regression function and the
riesz
riesz regression
we formulate estimation of the nuisance parameters the regression function and the riesz representer as minimizing the discrepancy between neyman orthogonal scores computed with known and unknown nuisance parameters which we refer to as neyman targeted estimation
mathematically annealed heterogeneity introduces a variance-weighted demographic noise term that penalizes across-environment fitness variance and effectively rescales the population size thereby biasing
evolution
evolutionary dynamics
mathematically annealed heterogeneity introduces a variance-weighted demographic noise term that penalizes across-environment fitness variance and effectively rescales the population size thereby biasing evolution toward generalist solutions
urban science has largely relied on universal models rendering the heterogeneous and locally specific nature of
cities
large cities
urban science has largely relied on universal models rendering the heterogeneous and locally specific nature of cities effectively invisible
further when there is variation in information gathering abilities our model implies that the sharing of space between smaller subgroups of the population rather than the whole population is optimal for
collective
collective systems
further when there is variation in information gathering abilities our model implies that the sharing of space between smaller subgroups of the population rather than the whole population is optimal for collective knowledge sharing
an alternative hypothesis is that discrimination is supported by lossy compression of visual inputs efficiently coding
sensory
receptive fields
an alternative hypothesis is that discrimination is supported by lossy compression of visual inputs efficiently coding sensory information by discarding seemingly irrelevant details
our approach enables measuring the electrical characteristics of the same flake before and after intercalation permitting us to precisely identify the effect of the guest species on the tas2
transport
transport properties
our approach enables measuring the electrical characteristics of the same flake before and after intercalation permitting us to precisely identify the effect of the guest species on the tas2 transport properties
redirected walking utilizes gain adjustments within perceptual thresholds to allow natural navigation in large scale
virtual
virtual reality
redirected walking utilizes gain adjustments within perceptual thresholds to allow natural navigation in large scale virtual environments within confined physical environments
in addition active learning which aims to proactively select the most informative unlabeled samples for annotation has been explored in semi-supervised 3d
action
action recognition
in addition active learning which aims to proactively select the most informative unlabeled samples for annotation has been explored in semi-supervised 3d action recognition for training sample selection
in order to understand how galaxy quenching is related to galaxy sizes we performed a demographic study of 46 massive
quiescent
host galaxy
in order to understand how galaxy quenching is related to galaxy sizes we performed a demographic study of 46 massive quiescent central galaxies with stellar mass from 10 10
further we found high agreement among five proxy llms while each individual
llm
llm reasoning
further we found high agreement among five proxy llms while each individual llm had low correlation with users evaluations
these findings support a clear takeaway improving representation learning is a direct and useful path to robust
world
world models
these findings support a clear takeaway improving representation learning is a direct and useful path to robust world models delivering reliable long-horizon predictions without enlarging the dynamics module
a h 13 co 3--2 velocity gradient of 29 km s -1 pc -1 is evident along the
filament
magnetic field
a h 13 co 3--2 velocity gradient of 29 km s -1 pc -1 is evident along the filament s long axis aligned with the magnetic field direction
furthermore we develop an unsupervised online learning approach that can learn this model on a single-trial basis building its vocabulary incrementally as it is exposed to new tasks and inferring the
latent
recurrent neural networks
furthermore we develop an unsupervised online learning approach that can learn this model on a single-trial basis building its vocabulary incrementally as it is exposed to new tasks and inferring the latent epoch structure as a time-varying computational context within a trial
the most decisive among these properties is the
convexity
strongly convex
the most decisive among these properties is the convexity or non-convexity of the loss function
the tree supports text polylog n -time operations and requires a static lookup table of size text poly n text polylog u -- but in exchange for these the
tree
tree edit
the tree supports text polylog n -time operations and requires a static lookup table of size text poly n text polylog u -- but in exchange for these the tree is able to achieve a remarkable space guarantee
through this work we study how data filtering and
contamination
synthetic data
through this work we study how data filtering and contamination interact to shape both benchmark and generative performance
additionally our model offers important insight on social phenomena such as false levels of support among cooperators often observed in agreement negotiations instances of non-strict
consensus
collective action
additionally our model offers important insight on social phenomena such as false levels of support among cooperators often observed in agreement negotiations instances of non-strict consensus when two people support the same political position albeit with different intensities and competitive situations such as in competitions with disproportionate profit and losses
to address these challenges we present spade sparsity adaptive depth estimator a monocular depth
estimation
depth estimation
to address these challenges we present spade sparsity adaptive depth estimator a monocular depth estimation pipeline that combines pre-trained relative depth estimator with sparse depth priors to produce dense metric scale depth maps
from a theoretical perspective we formally show that the aggregation of individual components forecasted with pooled
panel
panel data
from a theoretical perspective we formally show that the aggregation of individual components forecasted with pooled panel data regressions is superior to direct aggregate forecasting due to lower estimation error
we apply this theory to identify the geometric mechanisms of third-order
nonlinear
third-order nonlinear transport
we apply this theory to identify the geometric mechanisms of third-order nonlinear transport in materials both with and without time-reversal symmetry such as 2d materials topological materials and altermagnets
we design and demonstrate heuristic quantum
advantage
quantum batteries
we design and demonstrate heuristic quantum advantage with peaked circuits hqap circuits on quantinuum s system model h2 quantum processor
bidirectional yet asymmetric causality between urban systems and traffic dynamics in 30
cities
large cities
bidirectional yet asymmetric causality between urban systems and traffic dynamics in 30 cities worldwide
in addition the theoretical results are further illustrated by several
numerical
numerical simulations
in addition the theoretical results are further illustrated by several numerical simulations
in contrast both problems admit algorithms against oblivious
adversaries
time complexity
in contrast both problems admit algorithms against oblivious adversaries that achieve operatorname polylog n amortized update time behnezhad derakhshan hajiaghayi stein sudan focs 19
our work not only advances the state-of-the-art in
avsr
achieves state-of-the-art
our work not only advances the state-of-the-art in avsr but also extends its core components to multiple frameworks for diverse scenarios
secondly sufficient conditions for a central
limit
asymptotic normality
secondly sufficient conditions for a central limit theorem with a standard rate of convergence are supplied
to complement this we derive a tight upper bound of 2n k-1 for chordal bigraphs and 54n k-1 for grid intersection graphs gig a prominent graph class residing in four
ferrers
ferrers dimension
to complement this we derive a tight upper bound of 2n k-1 for chordal bigraphs and 54n k-1 for grid intersection graphs gig a prominent graph class residing in four ferrers dimensions and capturing planar bipartite graphs as well as bipartite intersection graphs of rectangles
to address these gaps we propose a hybrid
visual
visual navigation
to address these gaps we propose a hybrid visual tracking framework that bridges advanced perception with real-time servo control
motivated by the quadratic nature of self-attention we hypothesize that vits represent whether two patches belong to the same
object
human cognition
motivated by the quadratic nature of self-attention we hypothesize that vits represent whether two patches belong to the same object a property we term issameobject
augmented reality ar technologies are redefining how we perceive and interact with the world by seamlessly integrating digital elements into our
physical
mobile ar
augmented reality ar technologies are redefining how we perceive and interact with the world by seamlessly integrating digital elements into our physical surroundings
unmanned aerial vehicles with their airborne full-sample continuous trajectory observation
bring
traffic states
unmanned aerial vehicles with their airborne full-sample continuous trajectory observation bring new opportunities for macro- and micro-traffic state estimation
score-based constrained generative modeling via langevin diffusions with
boundary
generative models
score-based constrained generative modeling via langevin diffusions with boundary conditions
through this self-play core drones continuously compete against increasingly proficient versions of themselves naturally escalating the difficulty of
competitive
multi-drone racing
through this self-play core drones continuously compete against increasingly proficient versions of themselves naturally escalating the difficulty of competitive interactions
to cope with the decoding issue an interference-canceling ic first
decoding
wireless systems
to cope with the decoding issue an interference-canceling ic first decoding strategy is proposed at the access point ap which can partially tackles collision problems contributing to a higher system throughput
therefore in this work the options framework is applied and tailored to
autonomous
collision avoidance
therefore in this work the options framework is applied and tailored to autonomous driving tasks on highways
unfortunately in too many cases today s ai is not accountable -- we cannot
question
ai use
unfortunately in too many cases today s ai is not accountable -- we cannot question it enter into a discussion with it let alone sanction it
anisotropic hot carrier relaxation and coherent
phonon
phonon polaritons
anisotropic hot carrier relaxation and coherent phonon dynamics in type-ii weyl semimetal tairte4
we aimed to evaluate the performance of these models under
conditional
conditional exchangeability
we aimed to evaluate the performance of these models under conditional exchangeability violations and the utility of negative control outcomes ncos as a diagnostic
we quantify the information within the illumination patterns using the singular value decomposition svd and reconstruct reflectance spectra specifically hemoglobin and several green vegetation
spectra
spectral range
we quantify the information within the illumination patterns using the singular value decomposition svd and reconstruct reflectance spectra specifically hemoglobin and several green vegetation spectra using the pseudoinverse of the svd for a given amount of noise
many causal estimands such as average treatment effects under
unconfoundedness
causal inference
many causal estimands such as average treatment effects under unconfoundedness can be written as continuous linear functionals of an unknown regression function
towards quadrupedal jumping and walking for dynamic locomotion using
reinforcement
reinforcement learning
towards quadrupedal jumping and walking for dynamic locomotion using reinforcement learning
such costs can however be reduced thanks to effective
travel
travel information
such costs can however be reduced thanks to effective travel information strategies during traffic disruptions
these include the two spike train distances isi- and spike-distance as well as the coincidence detector spike-synchronization and its
directional
spike train
these include the two spike train distances isi- and spike-distance as well as the coincidence detector spike-synchronization and its directional companion spike-order
first to ensure safety at the path-segment level a segment-wise conservative collision test is applied where risk-prone trajectory path segments are recursively subdivided until
collision
obstacle avoidance
first to ensure safety at the path-segment level a segment-wise conservative collision test is applied where risk-prone trajectory path segments are recursively subdivided until collision risks are eliminated
our methods and findings offer not only practical applications for quantum networks but also lead to a deeper understanding of
multipartite
genuine multipartite
our methods and findings offer not only practical applications for quantum networks but also lead to a deeper understanding of multipartite entanglement structures
specifically we propose an additive instrumental variable framework to identify mean potential outcomes and the average
treatment
average treatment effect
specifically we propose an additive instrumental variable framework to identify mean potential outcomes and the average treatment effect with a weighting function
motivated by practical applications we use this baseline to develop a new framework for fast approximate matrix
multiplication
matrix multiplication
motivated by practical applications we use this baseline to develop a new framework for fast approximate matrix multiplication amm via low-degree approximations of the cksu polynomials
numerical results show that our proposed framework can achieve around 60-90 performance gains over its time division duplex tdd where the
uplink
uplink communication
numerical results show that our proposed framework can achieve around 60-90 performance gains over its time division duplex tdd where the uplink and downlink transmissions are served in different orthogonal time slots
to answer the first question we introduce a fisher-information-based characteristic function of the
cooperative
cooperative game
to answer the first question we introduce a fisher-information-based characteristic function of the cooperative game which yields probability distributions that generate coordinated spatio-temporal patterns
our multilayer structure supports hybrid plasmonic
waveguide
waveguide modes
our multilayer structure supports hybrid plasmonic waveguide modes that can manifest as various orders of quasiparticles
subsequent data-driven discovery methods have sought to address this limitation but are often limited by simple typically linear
encoding
encoding models
subsequent data-driven discovery methods have sought to address this limitation but are often limited by simple typically linear encoding models
in this paper we unify the perspectives of stochastic processes and
reinforcement
policy learning
in this paper we unify the perspectives of stochastic processes and reinforcement learning through action-driven processes and illustrate their application to spiking neural networks
this progressive learning journey guides agents from mastering fundamental flight control to executing sophisticated cooperative
multi-drone
deep reinforcement
this progressive learning journey guides agents from mastering fundamental flight control to executing sophisticated cooperative multi-drone racing strategies
recent work has shown that different large language models llms converge to similar and accurate input
embedding
representation learning
recent work has shown that different large language models llms converge to similar and accurate input embedding representations for numbers
however their success is tied to one core capability reliable object
detection
image fusion
however their success is tied to one core capability reliable object detection in complex and multimodal environments
in the literature of cognitive neuroscience researchers tend to assume a linear
relationship
task performance
in the literature of cognitive neuroscience researchers tend to assume a linear relationship between brain activation level and task performance however controversial findings have been reported in participants at different ages and different proficiency levels
reinforcement learning rl is widely used to produce robust robotic
manipulation
robotic manipulation
reinforcement learning rl is widely used to produce robust robotic manipulation policies but fine-tuning vision-language-action vla models with rl can be unstable due to inaccurate value estimates and sparse supervision at intermediate steps
these results and the simplicity of the hcf-based system pave the way to a high-performance and scalable solution for ultra-stable
laser
pulsed laser
these results and the simplicity of the hcf-based system pave the way to a high-performance and scalable solution for ultra-stable laser sources
overall they are not yet reliable as standalone zero-shot reasoners but exhibit encouraging signs as complementary
visual
reasoning tasks
overall they are not yet reliable as standalone zero-shot reasoners but exhibit encouraging signs as complementary visual engines alongside dedicated reasoning models
modern llms are trained to think primarily via explicit text generation such as chain-of-thought
cot
thinking traces
modern llms are trained to think primarily via explicit text generation such as chain-of-thought cot which defers reasoning to post-training and under-leverages pre-training data
this note introduces a unified theory for
causal
causal inference
this note introduces a unified theory for causal inference that integrates riesz regression covariate balancing density-ratio estimation dre targeted maximum likelihood estimation tmle and the matching estimator in average treatment effect ate estimation
to address this we leverage a suite of representational similarity metrics-each capturing a distinct facet of representational correspondence such as geometry unit-level tuning or linear decodability-and assess brain region or model separability using
multiple
human brain
to address this we leverage a suite of representational similarity metrics-each capturing a distinct facet of representational correspondence such as geometry unit-level tuning or linear decodability-and assess brain region or model separability using multiple complementary measures
in particular we show the possibility of detecting einstein-podolsky-rosen steering and mode entanglement of non-gaussian
states
multipartite entanglement
in particular we show the possibility of detecting einstein-podolsky-rosen steering and mode entanglement of non-gaussian states from linear measurements only
unlike methods based on reinforcement learning or imitation
learning
reinforcement learning
unlike methods based on reinforcement learning or imitation learning which require specified rewards or labeled expert objectives our gan-based architecture learns directly from the joint distribution of observed holdings and market data
deploying machine learning models on compute-constrained devices has become a key building block of modern
iot
machine learning
deploying machine learning models on compute-constrained devices has become a key building block of modern iot applications
normative reasoning is a type of reasoning that involves
normative
reasoning curriculum
normative reasoning is a type of reasoning that involves normative or deontic modality such as obligation and permission
finally we derive electronic structure descriptions from a series of representative chemical models and compute the energetics from quantum experiments on
quantinuum
quantum batteries
finally we derive electronic structure descriptions from a series of representative chemical models and compute the energetics from quantum experiments on quantinuum s h-series ion trap devices using up to 19 qubits
our work thus establishes a computationally efficient paradigm for probabilistic forecasting stratospheric anomalies and showcases
generative
generative ai
our work thus establishes a computationally efficient paradigm for probabilistic forecasting stratospheric anomalies and showcases generative ai s potential to deepen the physical understanding of atmosphere-climate dynamics
these results suggest that llms can serve as practical proxies for otherwise unrevealed human thinking traces enabling label-only corpora to be extended into thinking-trace-augmented resources that enhance the reliability of
llm
llm responses
these results suggest that llms can serve as practical proxies for otherwise unrevealed human thinking traces enabling label-only corpora to be extended into thinking-trace-augmented resources that enhance the reliability of llm raters
our approach learns from the latent embeddings of paired prompts encoding target and converse behaviors to dynamically adjust activations connecting the
language
vision-language models
our approach learns from the latent embeddings of paired prompts encoding target and converse behaviors to dynamically adjust activations connecting the language modality with image context
the results demonstrate that the agent s training against opponents enables deliberate overtaking behaviours with an
overtaking
wheel-to-wheel racing
the results demonstrate that the agent s training against opponents enables deliberate overtaking behaviours with an overtaking rate of 87 compared 56 for an agent trained just to race
in the static retrieval problem a data structure must answer
retrieval
data structure
in the static retrieval problem a data structure must answer retrieval queries mapping a set of n keys in a universe to v -bit values
here we study continual learning and the compositional
reuse
continual learning
here we study continual learning and the compositional reuse of learned computations in recurrent neural network rnn models using a novel two-system approach one system that infers what computation to perform and one that implements how to perform it
this research develops a composite liveability index for greater london based on metrics related to the proximity density and diversity of pois along with population density and investigates how neighbourhood liveability relates to active
travel
travel information
this research develops a composite liveability index for greater london based on metrics related to the proximity density and diversity of pois along with population density and investigates how neighbourhood liveability relates to active travel behaviour
the coordination of multiple autonomous agents in high-speed competitive
environments
autonomous driving
the coordination of multiple autonomous agents in high-speed competitive environments represents a significant engineering challenge
we optimize the cavity-waveguide interface to minimize
photon
photonic circuits
we optimize the cavity-waveguide interface to minimize photon losses and demonstrate that its adjustment allows precise tuning of the light-matter interaction