Your AI research companion

1. cs.LGMAGE: Towards Generalizable Multi-timescale EEG RepresentationsTuan-Kiet Doan, Van-Chien Nguyen, Trung-Hieu Tran, Quang-Hung Pham, Ngoc-Son Vu…

   The authors introduce an EEG model that can learn from brain signals over several time ranges and choose which range matters most for a given input. Across seven datasets, it performs better than existing EEG foundation models while using fewer …

2. cs.LGProcess then Retrieve: Investigating the Origins of the Modality Gap in Vision-Language ModelsArpita Shanbhag, Julia Tran, Dhruv Mandala, Ayda Sultan

   The study investigates why image and language features in vision-language models can be unevenly aligned. Its analysis suggests that text representations need more early processing than image representations, offering a possible explanation for this mismatch.

3. cs.CVMed-DualLoRA: Local Adaptation of Foundation Models for 3D Cardiac MRIJoan Perramon-Llussà, Amelia Jiménez-Sánchez, Grzegorz Skorupko, Fotis Avgousti…

   The authors develop a way for hospitals to adapt a large heart-MRI model together without sharing their private data. It keeps some model adjustments local to each hospital, which can improve fit to local data while reducing communication between sites.

Every weekday, the new papers in your categories, each with a short summary. See a whole paper alert →

This week’s preprints test whether AI improvements survive contact with new hospitals, financial markets, physical laws and the people expected to use them.

6 papers put into perspective

1. Tutors change their questions, but still dominate the conversation
2. A convincing heat map is not an error alarm
3. Physical laws give neural networks a close rival
4. Reading financial news is not the same as beating the market

1. astro-ph.GAA DOLPHOT/Euclid Study of the Resolved, Faint Outer Milky Way Globular Cluster AM-1Daniel R. Weisz, Alessandro Savino, Andrew E. Dolphin, Meredith J. Durbin, Abig…

   The authors built a tool to measure individual stars in Euclid images and used it to study the distant, faint cluster AM-1. Their deeper measurements show that AM-1 is an old, diffuse cluster and provide a method other researchers can …

2. astro-ph.GAFirst detection of suprathermal CH$^+$ rotational emissions from chemical pumping: JWST/MIRI observations of …David A. Neufeld, Joel D. Green, Dan M. Watson, Paule Sonnentrucker

   The researchers use James Webb Space Telescope observations to detect rotational light from CH+ ions formed in a shocked cloud of gas. The pattern of light reveals how the ions are created and gives clues about the energy involved in …

3. astro-ph.GAFirst detection of the forbidden rotational spectrum of H$_3^+$: JWST/MIRI observations of the shocked gas in…David A. Neufeld, Joel D. Green, Dan M. Watson, Paule Sonnentrucker

   Using the James Webb Space Telescope, the researchers detect a previously unseen set of rotational signals from H3+ in gas shocked by a young star’s outflow. The signals reveal conditions in hot molecular gas and provide a new way to …

Every weekday, the new papers in your categories, each with a short summary. See a whole paper alert →

This week’s preprints connect ambitious new observations with a quieter challenge: understanding how instruments, models and reused measurements shape what astronomers can claim.

6 papers put into perspective

1. Weighing the Milky Way means testing the scales
2. A black hole measured through the stars around it
3. A cold nearby world in a deep view of the Universe
4. A low-frequency signal with an unresolved source

1. quant-phExperimental visual quantum state discriminationA. Duplinskiy, M. J. Filipovich, A. I. Lvovsky

   The authors train a specially designed optical system to distinguish visual inputs by measuring incoming photons in a task-specific way. In tests, it made fewer classification errors than ordinary image capture, including in an experiment with very noisy photon measurements.

2. quant-phCan One Hear the Size of a Clock?Markus Baumann, Claudia Linnhoff-Popien

   The paper finds a rule connecting the number of states in a randomly switching system to how many sustained oscillations it can show. It also suggests that observed oscillations can help determine the minimum size of an otherwise hidden system.

3. quant-phSymplectic Reck: In-Situ Learning of Gaussian Quantum OperationsJanet F. Zhong, Renwen Yu, Paul-Alexis Mor, Charles Roques-Carmes, Aviv Karniel…

   The authors develop a photonic circuit that can learn to undo an unknown transformation of light by adjusting itself using measured feedback. Their design extends self-adjusting optical networks to a broader class of quantum operations and is tested numerically under …

Every weekday, the new papers in your categories, each with a short summary. See a whole paper alert →

This week’s preprints explore how to preserve quantum information, extend the reach of quantum measurements and distinguish impressive hardware from computational advantage.

5 papers put into perspective

1. A quantum memory that cools its own escape routes
2. Infrared fingerprints, read through different light
3. Keeping track of an almost-free mass
4. Why more detected photons need not mean harder computation

1. cond-mat.softHierarchical Design of Ambidirectional LCE Sheets for Three-State Gaussian Shape MorphsAtalaya Milan Wilborn, Gurminder K. Paink, Raphael Kay, Patryk Wasik, Friedrich…

   The researchers designed thin, flexible sheets that change from one curved shape to another as they warm, and showed that their color can change along with their shape. This could help create soft machines and surfaces that adapt in several …

2. cond-mat.softManufacturing Complex Structures Through Interfacial Self-OrganizationJacob Garrett, Abhinit Sati, Edward Chien, Emily Whiting, Valeri Frumkin

   The researchers made complex solid shapes by using a thin wire framework to guide liquids into the desired form, rather than building the shape point by point. The approach could make intricate designs, including branching structures and hollow forms, easier …

3. cond-mat.softRapid Antigen Detection Tests Empowered by Cholesteric Spherical Reflectors for Expanded CapabilitiesChih-Chieh Chien, Yan-Song Zhang, Jean-François Conrotte, Carole L. Linster, Si…

   The researchers developed reflective particles that can make rapid infection tests easier to read and help distinguish different diseases by color and reflected light. Their work suggests a way to detect smaller amounts of virus and test for several diseases …

Every weekday, the new papers in your categories, each with a short summary. See a whole paper alert →

This week’s preprints explore how matter changes under extreme conditions, how quantum motion can be reversed, and when an X-ray measurement begins to disturb the electrons it seeks to reveal.

6 papers put into perspective

1. Carbon beyond diamond, caught in a crystal pattern
2. Watching superconducting electrons leave as a pair
3. Charged disturbances travel through a liquid crystal
4. A quantum round trip needs the right timing

1. hep-phProof of multiloop high-multiplicity integrand reduction coefficients in the 't Hooft-Veltman schemeShengli Yi

   The researchers gave two independent proofs of a formula used to simplify calculations involving many-particle interactions in quantum field theory. They also explain what happens when the usual mathematical basis becomes singular but the full expression remains finite.

2. gr-qcAnalogue Gravity: the Subcritical CaseF. Belgiorno, S. L. Cacciatori, A. Viganò

   The researchers examined whether waves can still be created in a model of flowing media even when there is no horizon, a boundary often associated with this effect. They show that creation can occur in some conditions, but not in …

3. gr-qcObservational signatures of Bardeen black holes under nonlinear electrodynamicsHaiyuan Feng, Dan-Dan Cui

   The researchers calculated how a Bardeen black hole would look if its light follows paths affected by nonlinear electromagnetic fields. They find that magnetic charge changes the apparent shadow and bright rings, offering possible clues for distinguishing this model from …

Every weekday, the new papers in your categories, each with a short summary. See a whole paper alert →

From elusive particle decays to tractable models of quantum gravity, this week’s selection shows how precision measurements and carefully constrained calculations make hidden physics accessible.

6 papers put into perspective

1. A rare decay comes into view
2. A charm quark keeps its quiet cone
3. Revisiting a key ingredient of the muon’s magnetism
4. Mapping spin in swirling nuclear matter

1. q-bio.NCSpike-to-Field Mechanisms of Turbulence-Like Dynamics in Spatial Spiking Neural NetworksZequan Liu

   The authors built a simulated network of spiking neurons to investigate how brain-like patterns of activity might arise from individual neural signals. They found that timing, signal spread, and dendritic effects all matter, so large-scale patterns should not be interpreted …

2. q-bio.NCLanguage processing in the autistic brain: A meta-analysis of neuroimaging studies across modalities and ageHanlin Wu, Bei Xiao, Qingxi He

   The authors combined findings from brain-imaging studies of language in autistic and non-autistic people. Both groups showed a similar broad language network, but autistic participants showed weaker activity in parts of the left side and stronger activity in one right-side …

3. q-bio.NCFrom MNH to the Friedman--Komogortsev Method---which changes improve eye-movement classification?Lee Friedman, Mehedi Hasan Raju, Oleg Komogortsev

   The authors tested which parts of an eye-movement classifier actually improve its agreement with trained human reviewers. They found that correcting a small timing delay explains the main improvement, so users of the older method can make that correction directly.

Every weekday, the new papers in your categories, each with a short summary. See a whole paper alert →

This week’s preprints explore durable memories, signals of approaching disease, the reliability of consciousness tests, and unexpected links between genes, glia, pain and experience.

6 papers put into perspective

1. A changing hippocampus can still remember
2. A molecular warning before prion symptoms
3. When a consciousness test finds commands that were never given
4. Sleep’s genetic link to the brain’s insulation

1. medRxiv: Infectious DiseasesMolecular typing and characterisation of antimicrobial resistance determinants of Neisseria gonorrhoeae among…M. M. Mdingi, B. Kullin, H. Jung, N. Mitchev, R. P. Peters

   Researchers tested ways to identify gonorrhoea strains and detect drug-resistance markers directly from patient samples in South Africa. The tests agreed well with genome sequencing, but some resistant infections lacked the markers, so laboratory susceptibility testing remains important.

2. q-bio.BMOriDynamics: Towards a World Model for Biomolecular DynamicsJiahao Yang, Dong Liu, Hongliang Li, Ji Chen, Yuehua Liu, Daoyun Ding et al.

   The authors developed a method for generating many possible shapes of proteins and their interacting partners, including how their atoms move together. It outperformed the tested alternatives and could recover different functional shapes and changes caused by binding partners.

3. q-bio.GNPublications as benchmarks: auditing AI agents on spatial transcriptomics alignmentYiqun T. Chen, Jan Matthias, Jianing Yao, Hongyu Eric Zhao, Stephanie C. Hicks

   The authors turned findings in scientific papers into tests that can be used to check how well AI agents perform specialized research tasks. In tests involving spatial gene-activity maps, the best agents matched or beat existing methods on most tasks, …

Every weekday, the new papers in your categories, each with a short summary. See a whole paper alert →

This week’s preprints explore how immune protection travels between generations, how microbes keep time and move genes, and how genomic evidence can overturn familiar histories.

6 papers put into perspective

1. A maternal legacy that limits damage rather than infection
2. The cattle family tree gets a different root
3. A cellular doorway for Langya virus
4. A familiar bacterium keeps its own time

1. physics.ao-phRescene-ESM: an Earth system model built around a frozen AI weather modelMinjong Cheon, Kaoru Kakinuma

   The authors present a climate model that combines a frozen AI weather system with components representing the ocean, land, rivers, and other parts of Earth. It runs quickly, reproduces key observed climate behaviour, and can trace how emissions affect both …

2. physics.ao-phGuiding future global temperature distributions with simulations and observationsFrancesco Immorlano, Elijah Tavares, Felix Draxler, Juan Nathaniel, Violette La…

   The researchers built a machine-learning tool that generates realistic patterns of daily temperatures while adjusting climate-model projections toward observed weather. It reduced errors in projected extremes and can produce large sets of possible futures without running more costly climate simulations.

3. physics.ao-phNew Astronomical, Meteorological and Geological Study of the "Torre Spaccata District" - VII district of the …Diego Tasselli, Stefano Ricci, Pamela Bianchi

   This study examined decades of weather and environmental data from a district of Rome, focusing on how much heating is needed in winter. Heating demand fell sharply in recent decades, while year-to-year variation increased, reflecting long-term climate change and the …

Every weekday, the new papers in your categories, each with a short summary. See a whole paper alert →

This week’s preprints trace how overlooked nutrients, human ignitions, fractured rock and computing choices can change our understanding of environments on Earth and beyond.

5 papers put into perspective

1. Antarctic plankton need more than iron
2. People start desert fires; plants help them spread
3. Listening for damage inside a mountainside
4. Why moonquakes keep ringing

1. math.PRA Geometric Construction of the Radial Dunkl and Heckman-Opdam ProcessesJake Mundo

   This work constructs a broad family of random motions from the geometry of symmetric spaces, including processes related to Bessel processes and Dyson Brownian motion. It also shows that, in a limiting case, these random motions become the smooth geometric …

2. math.OCA Hierarchical Variational Inequality Approach to Generalized Nash Equilibrium Seeking with Nonlinear Shared …Tatiana Tatarenko, Puya Latafat, Alberto Bemporad

   The authors develop algorithms for finding stable outcomes in games where players’ choices are linked by shared nonlinear constraints. Their methods work with local information and projections, and they prove rates at which the computed outcomes approach equilibrium.

3. math.PRThe satisfiability conjecture for random k-SAT formulas. Human-friendly exposition of the recent OpenAI resol…David Gamarnik

   This note gives a detailed, accessible explanation of a recent proof about when random collections of logical clauses stop being satisfiable as more clauses are added. The result resolves a long-standing question, and the note aims to make the proof …

Every weekday, the new papers in your categories, each with a short summary. See a whole paper alert →

New preprints tackle magnetic correlations, statistical approximation and the geometry of efficient representations, revealing both sharp limits and unexpected room for improvement.

6 papers put into perspective

1. A magnetic prediction meets a probabilistic proof
2. The best lattice for rounding space
3. Normal approximation without a dimension-dependent constant
4. How evenly can points be divided?

1. cs.SEHow Much Skill Orchestration Can LLM Agents Compile Away? Execution-Grounded Specialization for Token-Efficie…Jiayi Ni, Zac Wang, Alex Zhang, Zheng Liu, Roopa Prabhu

   The authors study whether an AI agent can learn to reuse a completed workflow instead of rebuilding it from instructions each time. Their system replays a saved, adaptable version of a successful run, cutting token use substantially while maintaining comparable …

2. cs.CRzk-DEAP: A Zero-Knowledge Decentralized Encrypted Aggregation ProtocolJude Ramanan, Georgiana Haldeman, Noah Apthorpe

   This paper presents a way to combine private inputs while checking that they obey rules, without relying on a trusted central party. Its cryptographic checks tie each proof to the value actually being encrypted, preventing someone from substituting a different …

3. cs.SEMicius-Flash: Lightweight Agentic Intelligence for EngineeringHongcheng Gao, Weiheng Li, Yipeng Wei, Yu Lei, Shuo Sun, Yi Li et al.

   The authors introduce Micius-Flash, an AI model designed for engineering tasks, and train it using specialist models and examples it generates itself. It outperformed a same-size baseline on hardware coding and semiconductor tasks, and demonstrations show it using engineering tools …

Every weekday, the new papers in your categories, each with a short summary. See a whole paper alert →

New preprints explore the information hidden in aircraft noise, the limits of quantum and economic advantage, and why apparently good software and energy forecasts can disappoint in practice.

6 papers put into perspective

1. An aircraft overhead can become a location clue
2. Security scanners can improve their scores without improving discrimination
3. A proposed boundary on unstructured quantum advantage
4. For the grid, uncertainty needs a timeline

1. econ.GNHave wind and solar energy increased electricity price volatility? And what is electricity price volatility, …Lion Hirth

   The paper distinguishes price differences between delivery periods from price changes for the same delivery period as markets receive new information. Using a model and 25 years of German data, it finds that the first measure fell and then rose, …

2. econ.EMGood Looks and the Academic Job Market: Measuring Attractiveness with Vision EmbeddingsFabian Slonimczyk, Shergazy Kimsanbaev

   The paper creates a computer-based way to estimate facial attractiveness using existing image models, then checks it against human ratings and uses it to study economics job applicants. Higher scores are linked to more favorable and longer recommendation letters, especially …

3. econ.EMAsymptotic Theory of Integrated Pairwise-Difference Functionals of Kernel Estimators, with Applications to Tr…Julius Owusu

   The authors develop statistical tests for whether the effect of a treatment varies across people or groups, including tests of how much effects differ in a particular direction. Simulations support the method, and an example examines how 401(k) eligibility relates …

Every weekday, the new papers in your categories, each with a short summary. See a whole paper alert →

New preprints explore how infrastructure costs, industrial resilience, AI access and public attention reshape the choices available to households, firms and governments.

6 papers put into perspective

1. When free water connections make economic sense
2. After floods, carmakers add backup rather than retreat
3. An AI advantage for software defenders
4. Public attention can turn violence into a call for help