Peer Revenue Publish or profit

Papers

Fresh from arXiv, typeset properly. Highlighted words trade on the market.

All of arXiv

Thursday 3 September

Computer Vision

Principia: Relational Physics Tests for Video Models

Evaluating physical reasoning in video models is difficult because absolute motion measurements depend on frame rate, object scale, and camera calibration, all of which are often ambiguous or unavailable in generated video. We propose a different approach. When two objects in the same scene obey the same physical law, their motions must satisfy predictable relationships, and these relationships hold independent of calibration. We introduce Principia, a benchmark that evaluates Newtonian physics through relational consistency between paired objects. Principia spans eight phenomena - gravity, restitution, friction, rotational inertia, projectile motion, momentum, pendulum, and mass-spring oscillation - across translational, rotational, collisional, and oscillatory dynamics, using real-world scenes recorded under controlled protocols. We also introduce a calibration-independent consistency score that quantifies physical violation directly in image space. Across thousands of generations from six state-of-the-art video generators, no model exceeds 0.42 on Principia despite all scoring around 0.8 on VBench. Vision-language models are evaluated on their ability to detect relational physics violations, with the best model achieving only 67% accuracy and most performing near chance level.
Read the full abstract

All authors Varun Varma Thozhiyoor, Shivam Tripathi, Venkatesh Babu Radhakrishnan, Anand Bhattad

Computation and Language

Compile by Training: Turning Natural-Language Specifications into Local Neural Functions

Many recurring text functions are easy to describe but difficult to implement with rules, while calling a large remote model for every input introduces repeated cost, latency, and dependency on a provider. We present compile by training, which turns a natural-language specification into a reusable neural function. At compile time, teacher models generate task-specific examples that are used to train a small adapter for a compact interpreter. The resulting function runs without the teachers and can be stored, versioned, and composed like ordinary software. On FuzzyBench-Hard, a subset on which the Program-as-Weights fast compiler produced no exact matches, compile by training reaches 83.6% semantic accuracy. This higher accuracy comes with a higher compile-time cost: roughly a minute rather than seconds for the fast compiler. We deploy the compiler in a public interactive service and demonstrate compiled functions in a multi-site website helper, a language-controlled 3D avatar, and a bidirectional English-Claudish translator.
Read the full abstract

Artificial Intelligence

Clean Engineering, Unstable Measurement: A Preregistered Reliability Failure of Black-Box LLM Observers on Shared Endpoints

Language-model judges now gate training data, score generations, and drive leaderboards. The judge is then a measurement instrument, resting on one rarely stated assumption: the same request, sent to the same model name, reads the same tomorrow. We audited that assumption in two preregistered campaigns with every threshold fixed in advance; neither got past validating its instrument. Across 52,988 audited request attempts, same-window repeat rankings agreed at Spearman 0.400 against a required 0.90, and byte-identical next-day replays agreed at 0.78 against a required 0.99, each time with the execution record at ceiling. Three mechanisms explain the gap: a label-to-meaning mapping that biased readouts as strongly as the signal; candidate gaps seven orders of magnitude below the instrument's own noise floor; and byte-identical inputs returning different rankings, a noise that exact-permutation readouts compound. Neither metric substitution nor sampling repaired it on the tested grid. Preregistered follow-ups bound the problem: waiting did not help on the days sampled (0.805 versus 0.800, replicated over five further days); switching providers did not help (four providers share the floor, medians 0.74 to 0.88, predicted by none of the metadata fields they expose); self-hosting on batch-invariant kernels helped only while the server was quiet; and on constructed errors with known gaps, the readout's separation tracks error type, not size. We distill the evidence into a three-level snapshot-identity ladder, eight design rules, and a reporting checklist; a pilot at roughly 2% of the study's call volume would have exposed both unreachable gates in advance. All results concern externally measured behaviour on shared serving infrastructure. On a shared endpoint, a model name is not a frozen instrument; a preregistered evaluation must measure its instrument before freezing any gate on it.
Read the full abstract

Computation and Language

ESPO: Error-Structured Prompt Optimization via Diagnose, Diversify, and Stabilize

Evolutionary prompt optimizers such as GEPA suffer from prompt bloat: each iteration appends rules and caveats, producing prompts up to 3$\times$ longer yet no more accurate. We trace this to three deficiencies - incomplete error observation, limited search diversity, and unreliable selection - and propose ESPO (Error-Structured Prompt Optimization), which decomposes prompt optimization into three phases: Diagnose clusters all training errors into structural patterns in one round; Propose generates candidates via four complementary strategies with independent biases; Select applies bootstrap stability selection. On seven public NLP benchmarks - Tweet, MMLU, GSM8K, HotpotQA, ScoNe, HoVer, and PUPA - ESPO improves average accuracy by $+$3.76 pp over the state-of-the-art (74.67% vs 70.91% for GEPA), matching or exceeding GEPA on every dataset while producing prompts 47% shorter (1,004 vs 1,878 chars) and faster at inference. Cross-model experiments across four additional student models (Gemma 3 12B, Mistral 14B, Qwen3 32B, Claude Haiku 4.5) show ESPO yields the best average accuracy on every model tested, with the largest gap on Qwen3 GSM8K (15.00% $\to$ 91.40%). A generalization bound (Appendix) grounds each phase in a corresponding term of the test-time gap, and the ablation confirms a key prediction: adding diversity without bootstrap selection actually hurts performance ($-$1.20%).
Read the full abstract

All authors Lihao Liu, Peng Tang, Kunwar Yashraj Singh, Shabnam Ghadar

Computer Vision

Puffin-World: Scaling a Unified Multimodal Model with Native 3D World States

We propose Puffin-World, a unified multimodal architecture that integrates physical understanding, spatial simulation, and 3D world generation and reconstruction without relying on external offline modules. To reliably construct and interact with 3D worlds, our framework jointly models three native world states: physics (gravity field and latitude), geometry (depth), and appearance (image), together with a unified Omni-Camera representation that supports diverse tasks and flexible motions. Beyond modeling these states, we introduce a strategy for propagating physical dynamics across future frames. By grounding absolute camera properties in the real world, Puffin-World enables physically consistent and visually stable world generation. We further couple appearance and geometry within a single generative process, jointly synthesizing each future view and reconstructing its underlying geometry. This unified paradigm enables interleaved closed-loop applications requiring synergy across multiple tasks, including mimic and self-calibrated world exploration. To scale Puffin-World to complex scenarios, we construct Puffin-16M, comprising 15 million vision-language-camera triplets and 1 million trajectories featuring various and challenging motions. To foster further research in this area, we released the code, models, and datasets.
Read the full abstract

All authors Kang Liao, Yihang Luo, Xiao-Ming Wu, Linyi Jin, Size Wu, Chunyu Lin, Yao Zhao, Fei Wang, Wei Li, Chen Change Loy

Neurons and Cognition

From anatomy to computation: axonal delay dispersion shifts neurons from event to sequence detection

Cortical neurons fire sparsely – often fewer than one spike per sensory window – so a neuron's spike count in that window carries little information, and its spike timing can carry more. That conduction delays convert firing order into synchrony is long established. What governs which class of temporal feature a neuron detects – one volley of coincident input, or two in a particular order – has not been examined. We propose a delay-signature framework in which the axonal conduction delays converging on a dendritic branch or a small neuron's soma constitute a physical key: only input sequences whose spike-time differences the delays compensate arrive synchronously, and coincidence detection, a threshold on input summed within an inhibition-compressed window, converts that synchrony into an all-or-none output. First, in simulations of an integrator-neuron model, a single physical scalar – the dispersion of the delay set – moves a population from event to order-selective sequence detection. The transition is emergent under random delays and connectivity, and an exact calculation reproduces it: at narrow dispersion, sequence detectors do not exist, and the dispersion at which they overtake event detectors tracks the inter-event interval. This maps a computational distinction onto the anatomical one between myelinated and unmyelinated projections, making myelination a switch on what a neuron computes. Second, the same dispersion sets the code's limits: it bounds the longest codable interval and fixes an absolute timing tolerance of order a millisecond, the same at every dispersion. Third, that millisecond window and the conduction velocity of afferents inside cortex together set column diameter, and column anatomy predicts the window measured in mouse barrel cortex. One anatomically measurable parameter thus shapes what a neuron detects and sets the limits of what it can represent.
Read the full abstract

Computation and Language

Legibility is Not Interpretability: Comparing Judged and Actual Importance in Chain-Of-Thought Reasoning

Reasoning traces from chain-of-thought models appear to offer a legible window into how a model arrives at its answer. A growing body of work treats them as such, using LLM judges to diagnose errors, evaluate faithfulness, and provide step-level supervision via process reward models and generative critics. These practices rely on the text of a reasoning step carrying information about its functional role. But does the text actually encode information about which reasoning steps matter? We operationalize the importance of a reasoning step as its advantage: the change in expected reward, e.g., producing the correct final answer, from including that step, estimated via Monte Carlo rollouts. Basing ground truth on these estimates, we evaluate whether LLM judges can identify high-advantage steps and find that sufficiently capable LLMs can outperform a prevalence baseline but fall well short of a noise ceiling. Fine-tuning a model as a step-level critic yields strong improvement for incorrect responses but remains distant from ceiling for correct responses, suggesting that step importance is only partially recoverable from the text of the reasoning trace. Our findings contribute to a growing body of chain-of-thought faithfulness work that cautions against treating the legibility of reasoning traces as interpretability, especially with implications for process reward modeling.
Read the full abstract

All authors Kevin Du, Alexander Hoyle, Laura Ruis, Acyr Locatelli

Robotics

GIFT: Guided Intermediate Feature Training via Action-Oriented Structural Supervision for Robotic Manipulation

Vision-language pre-training and predictive world modeling provide robot policies with rich semantic and dynamic visual features, but their native action and visual-prediction objectives may omit critical physical and task structure while retaining control-irrelevant visual redundancy. We call this mismatch between visual richness and control utility the action-sufficiency gap. We investigate whether this gap can be bridged by guiding intermediate features to preserve three control-relevant structure in robotic manipulation: geometry governing motion feasibility, affordance encoding instruction-relevant entities, and goals grounding instructions in task-relevant regions. To this end, we present GIFT (Guided Intermediate Feature Training), an architecture-flexible framework for learning intermediate features that translates these structures into training-time constraints through geometry alignment, affordance prediction, and goal-region reconstruction. We instantiate GIFT in a Vision-Language-Action (VLA) policy, a direct-action World-Action Model (WAM), and an inverse-dynamics WAM while retaining each model's action formulation. Under zero-shot transfer to LIBERO-Plus, GIFT-VLA, GIFT-WAM-Fast, and GIFT-WAM-IDM outperform StarVLA-OFT, Fast-WAM, and Fast-WAM-IDM by 4.6, 12.6, and 5.2 points, reaching 79.6%, 72.6%, and 87.8%, respectively. On RoboCasa, the three GIFT variants reach 61.4%, 83.6%, and 82.3%, outperforming their counterparts by 12.6, 9.0, and 8.4 points, respectively. Together, these results establish learning functionally structured intermediate features as a reusable principle across model-specific action formulations, with especially large gains on articulated-object tasks and high-precision real-world manipulation under unseen visual and spatial perturbations. Project page: https://openphoenix-team.github.io/GIFT-pages.
Read the full abstract

All authors Yupeng Zheng, Xiang Li, Songen Gu, Yuhang Zheng, Shuai Tian, Weize Li, Linbo Wang, Chaoyue Li, Qichao Zhang, Haoran Li, Zhongpu Xia, Ya-Qin Zhang, Shuicheng Yan, Dongbin Zhao

Cosmology and Nongalactic Astrophysics

Euclid preparation. The shape of halo profiles in $\Lambda$CDM and non-standard cosmologies

We study the shape of three-dimensional and projected dark-matter halo profiles extracted from cosmological $N$-body simulations in $\Lambda$CDM and non-standard cosmologies, using the DUSTGRAIN-PF and DEMNUni suites. The models considered include massive neutrinos, $f(\mathcal{R})$ gravity, and dynamical dark energy. By comparing density, mass, velocity-dispersion, and excess-surface-density profiles up to $5\,r_{500{\rm c}}$, we quantify the differential imprint of non-standard physics on halo structure in view of Euclid cluster WL studies. Our main analysis is performed at $z=1.1$, a high-redshift regime where the weak-lensing signal-to-noise starts to degrade, providing a conservative stress test for detectability; for DUSTGRAIN-PF we additionally analyse $z=0.5$ and $z=0.3$ snapshots. In low-mass haloes ($M_{\rm 200c}<7\times10^{13}\,M_\odot$), $f(\mathcal{R})$ gravity produces deviations of order $10\,\%$ in projected and three-dimensional profiles, especially in the outskirts where screening is less efficient. Massive neutrinos partially reduce this signal, reflecting the competition between free streaming and fifth-force-enhanced growth. Dynamical dark energy and massive-neutrino cosmologies generally induce smaller, few-percent deviations, with the largest effects again found in low-mass haloes and at large radii. Under simplified assumptions for \Euclid WL, detecting such profile differences at $z=1.1$ requires stacks of $\sim10^5$ haloes, while a few thousands haloes may be sufficient at $z\lesssim0.5$. This further calls for the need of integrating such precise modelling of non-standard effects – along with other observational effects – in any likelihood involving Euclid WL masses to avoid non-negligible systematic biases. Concentration–mass relations show weaker cosmology dependence, typically at the $\sim5\,\%$ level. [...]
Read the full abstract

All authors Euclid Collaboration: L. Pizzuti, G. Y. Ferron, A. Ragagnin, A. M. C. Le Brun, P.-S. Corasaniti, T. Gayoux, G. Rácz, E. Altamura, Z. Sakr, C. Carbone, M. Baldi, C. Giocoli, T. Castro, F. Pace, J. E. Taylor, S. Borgani, O. Luongo, C. T. Mpetha, R. E. Angulo, B. Altieri, S. Andreon, N. Auricchio, C. Baccigalupi, S. Bardelli, P. Battaglia, A. Biviano, E. Branchini, M. Brescia, S. Camera, V. Capobianco, V. F. Cardone, J. Carretero, M. Castellano, G. Castignani, S. Cavuoti, K. C. Chambers, A. Cimatti, C. Colodro-Conde, G. Congedo, L. Conversi, Y. Copin, F. Courbin, H. M. Courtois, M. Cropper, H. Degaudenzi, G. De Lucia, H. Dole, F. Dubath, X. Dupac, S. Dusini, S. Escoffier, M. Farina, F. Faustini, S. Ferriol, F. Finelli, P. Fosalba, S. Fotopoulou, M. Frailis, E. Franceschi, M. Fumana, L. Gabarra, S. Galeotta, K. George, B. Gillis, J. Gracia-Carpio, A. Grazian, F. Grupp, S. V. H. Haugan, S. Hemmati, W. Holmes, F. Hormuth, A. Hornstrup, K. Jahnke, M. Jhabvala, B. Joachimi, S. Kermiche, A. Kiessling, B. Kubik, K. Kuijken, M. Kunz, H. Kurki-Suonio, S. Ligori, P. B. Lilje, V. Lindholm, I. Lloro, M. Magliocchetti, G. Mainetti, O. Mansutti, O. Marggraf, M. Martinelli, N. Martinet, F. Marulli, R. J. Massey, N. Mauri, S. Maurogordato, E. Medinaceli, S. Mei, M. Meneghetti, E. Merlin, G. Meylan, P. Monaco, A. Mora, M. Moresco, C. Moretti, L. Moscardini, E. Munari, C. Neissner, J. W. Nightingale, C. Padilla, S. Paltani, F. Pasian, V. Pettorino, A. Pezzotta, S. Pires, G. Polenta, M. Poncet, L. A. Popa, F. Raison, A. Renzi, J. Rhodes, G. Riccio, I. Risso, E. Romelli, M. Roncarelli, R. Saglia, D. Sapone, B. Sartoris, P. Schneider, T. Schrabback, A. Secroun, E. Sefusatti, E. Sihvola, P. Simon, C. Sirignano, G. Sirri, L. Stanco, J.-L. Starck, P. Tallada-Crespí, A. N. Taylor, I. Tereno, N. Tessore, S. Toft, R. Toledo-Moreo, F. Torradeflot, I. Tutusaus, E. A. Valentijn, J. Valiviita, T. Vassallo, Y. Wang, J. Weller, A. Zacchei, G. Zamorani, F. M. Zerbi, E. Zucca, M. Ballardini, P. Bergamini, A. Boucaud, E. Bozzo, C. Burigana, R. Cabanac, M. Calabrese, A. Cappi, F. Caro, J. A. Escartin Vigo, J. García-Bellido, T. Gasparetto, J. Macias-Perez, R. Maoli, R. B. Metcalf, M. Pöntinen, E. Sarpa, V. Scottez, M. Sereno, M. Tenti, M. Tucci, M. Viel, M. Wiesmann, J. A. Acevedo Barroso, Y. Akrami, I. T. Andika, S. Anselmi, M. Archidiacono, G. Aricò, F. Atrio-Barandela, M. Baes, L. Bazzanini, D. Bertacca, M. Bethermin, F. Beutler, A. Blanchard, L. Blot, H. Böhringer, M. L. Brown, S. Bruton, A. Calabro, B. Camacho Quevedo, C. S. Carvalho, F. Cogato, T. E. Collett, S. Conseil, A. R. Cooray, P. Corcho-Caballero, M. Costanzi, B. Csizi, O. Cucciati, T. de Boer, F. De Paolis, G. Desprez, A. Díaz-Sánchez, S. Di Domizio, J. M. Diego, V. Duret, Y. Fang, A. Farina, A. Finoguenov, F. Fontanot, A. Franco, Y. Fu, K. Ganga, R. Gavazzi, E. Gaztanaga, Z. Ghaffari, F. Giacomini, F. Gianotti, E. J. Gonzalez, G. Gozaliasl, A. Gruppuso, M. Guidi, C. M. Gutierrez, A. Hall, N. A. Hatch, C. Hernández-Monteagudo, H. Hildebrandt, J. Hjorth, J. J. E. Kajava, Y. Kang, V. Kansal, D. Karagiannis, J. Kim, C. C. Kirkpatrick, A. Kovács, I. Kovačić, K. Koyama, S. Kruk, M. C. Lam, M. Lattanzi, L. Legrand, M. Lembo, G. Leroy, G. F. Lesci, J. Lesgourgues, L. Linke, A. Manjón-García, F. Mannucci, F. R. Marleau, C. J. A. P. Martins, M. Migliaccio, M. Miluzio, G. Morgante, S. Nadathur, K. Naidoo, A. Navarro-Alsina, S. Nesseris, F. Oppizzi, D. Paoletti, F. Passalacqua, K. Paterson, L. Patrizii, C. Pattison, R. Paviot, A. Pisani, D. Potter, G. W. Pratt, S. Quai, M. Radovich, G. Rodighiero, W. Roster, S. Sacquegna, M. Sahlén, D. B. Sanders, A. Schneider, D. Sciotti, E. Sellentin, S. Serjeant, F. Shankar, L. C. Smith, J. G. Sorce, I. Szapudi, K. Tanidis, C. Tao, F. Tarsitano, G. Testera, R. Teyssier, S. Tosi, A. Troja, C. Uhlemann, C. Valieri, A. Venhola, D. Vergani, G. Verza, P. Vielzeuf, S. Vinciguerra, M. von Wietersheim-Kramsta, L. Wang, A. H. Wright, H. W. Yeung

High Energy Physics (Theory)

Wormhole-induced gravi-axion mass distributions and the second-harmonic crossover

We compute the distribution of the wormhole-induced gravi-axion curvature mass across baby-universe sectors when the first two harmonics are retained. For independent complex Gaussian coefficients, with an optional displacement of the first channel, an exact global-minimum condition gives the mass distribution and its threshold coefficient. At fixed nonzero second-harmonic width, the cumulative probability below mass $M$ scales as $M^5$. For a weak second harmonic, the unit-harmonic $M^4$ behavior is recovered in an intermediate low-mass range above the crossover scale. We also distinguish the actual relative mass-squared error from the probability that a phase-independent bound fails to certify the unit-harmonic approximation. In the common-background quadratic correlator with unit residues, the lower threshold remains zero and the two-harmonic tail produces an $s^{3/2}$ nonanalytic term in the Euclidean momentum squared $s$. The results are conditional on the chosen state and effective wormhole coefficients.
Read the full abstract

Computer Vision

One Editor, Many Edits: A Unified Training-Free Framework for Diverse Video Editing

Video editing spans diverse editing paradigms, yet achieving high-quality instruction-guided and subject-guided editing within a single unified framework remains challenging. We introduce EditVid, a training-free framework combining sparse causal memory for local coherence, correspondence-based post-attention token injection for long-range identity preservation, and soft latent blending for edit locality. The same framework supports instruction-guided and reference-guided edits, including style transfer, attribute modification, object insertion, part-level editing, and subject replacement. On FiVE, EditVid achieves 78.16 FiVE-Acc, compared with 58.95 for the strongest evaluated training-free baseline, while obtaining competitive results on IVEBench. A user study further shows a 51.8% overall preference for EditVid over 7 competing methods.
Read the full abstract

All authors Adheesh Sunil Juvekar, Onkar Kishor Susladkar, Kiet A. Nguyen, Muntasir Wahed, Nabeel Bashir, Xiaona Zhou, Tianjiao Yu, Vedant Shah, Ismini Lourentzou

Machine Learning

Robust PAC Learning of Concurrent Stochastic Games

We introduce the first Probably Approximately Correct (PAC) learning framework for general-sum concurrent stochastic games (CSGs) with transition uncertainty, while addressing the challenge of Nash equilibrium (NE) existence. Our algorithm maintains data-driven $L^1$ confidence sets over transition kernels and solves a robust CSG to compute a social-welfare optimal $\varepsilon$-NE, using a robust MDP-based exploration mechanism to drive joint state-action coverage. Crucially, we introduce a Nash margin characterisation that enables principled reasoning about equilibrium existence: the framework either returns an $\varepsilon$-approximate NE whose social-welfare value is $\varepsilon$-close to optimal, or provides a sound certificate that no exact NE exists. Under a minimum reachability condition $p_{\mathrm{reach}}>0$ over relevant state-action pairs, the algorithm terminates after a polynomial number of trajectory samples, with sample complexity $\widetilde{O}\left( {R_{\max}^2 H^4 |S|^2 |A| / (p_{\mathrm{reach}} \varepsilon^2)} \right)$. Empirical results on benchmark CSGs demonstrate near-optimal performance, correct handling of equilibrium (non-)existence, and sample complexity consistent with theory.
Read the full abstract

Optimization and Control

Computation of Strong Solutions to Stochastic Variational Inequalities

This paper studies the computation of strong solutions of monotone variational inequalities (VIs) with Lipschitz continuous operators. Building on the idea of accumulative regularization (AR), we develop a general framework for VIs, with particular emphasis on stochastic settings. Under unbiased stochastic oracles with uniformly bounded variance $ \sigma^2$, AR computes an approximate solution with expected operator residual bounded by $\varepsilon$ using at most $ \widetilde{\mathcal{O}}\left(\tfrac{LD_0}{\varepsilon}+\tfrac{ \sigma^2}{\varepsilon^2}(\log\tfrac{LD_0}{\varepsilon})^3\right) $ stochastic oracle calls, where $L$ is the Lipschitz constant and $D_0$ bounds the initial distance to the solution. This substantially improves the existing $\mathcal{O}( \sigma^2/\varepsilon^4)$ complexity for residual reduction and matches the lower bound up to logarithmic factors. For strongly monotone VIs, measured by the distance to the solution, AR achieves the optimal oracle complexity when the strong monotonicity modulus is known. By treating the problem as merely monotone, AR still achieves nearly optimal complexity without knowledge of this modulus. We further introduce a state-dependent noise model applicable to general monotone VIs with potentially nonunique solutions, extending state-dependent noise analysis beyond the strongly monotone setting. We apply these results to policy evaluation in reinforcement learning, where we show that estimators of the projected Bellman operator satisfy our state-dependent noise condition. To the best of our knowledge, the AR framework, the improved complexity under uniformly bounded noise, the state-dependent noise model and its guarantees, the adaptivity to an unknown strong monotonicity modulus, and the improved guarantees for FTD learning are all new in the VI literature.
Read the full abstract

Strongly Correlated Electrons

Wavefunctions for Anyon Superconductors

Anyon superconductivity arises from the condensation of mobile anyons rather than from a conventional Cooper instability, yet a systematic wavefunction description remains lacking. We develop a hierarchy-wavefunction construction for superconducting states derived from parent topological orders and identify their off-diagonal long-range order, condensate charge, chiral central charge, and residual topological order through the plasma analogy and topological field theories. We construct Abelian and non-Abelian examples descending from the semion state, a $\nu=2/3$ hierarchy state, the $\nu=1/3$ Laughlin state, $\nu=1$ integer quantum Hall state, and Pfaffian state. Remarkably, for the semion case, the superconducting many-semion wavefunction is equivalent to a state of fermionized anyons filling two effective Landau levels, recovering Laughlin's original construction of semion superconductor. Finally, we show that hierarchy wavefunctions emerge naturally in the dilute, long-distance limit of the anyon-Hilbert-space formulation, which applies to ideal Chern bands and moiré bands of twisted bilayer MoTe$_2$. Our results establish a unified wavefunction-level framework linking anyon condensation, superconducting order, and topological field theory.
Read the full abstract

High Energy Physics (Phenomenology)

Axion Portal Dark Matter and the LUX-ZEPLIN High-Recoil Event

LUX-ZEPLIN (LZ) has reported one event compatible with a $248~\mathrm{keV}$ nuclear recoil. While initial attention has focused on inelastic dark matter, that interpretation depends sensitively on the uncertain high-speed tail of the Galactic halo. We develop instead an elastic realisation of the pseudoscalar-pseudoscalar interaction $L_4=(\bar\chi i\gamma^5\chi)(\bar N i\gamma^5N)$, whose contact templates give one of the largest local significances identified by LZ. The minimal model presented involves Dirac fermion dark matter and a pseudoscalar mediator. The pseudoscalar arises from a complex singlet and the minimal gauge-invariant UV completion also requires a single multi-TeV vector-like heavy quark. A representative model which realises thermal freeze-out has $m_\chi\sim 400~{\rm GeV}$ and $m_a\sim 1~{\rm GeV}$. The simplest model breaks from the isoscalar/isovector assumption, and the recoil spectrum lies between LZ's elastic $O_4^v$ and contact $L_4^v$ templates, suggesting a local significance of $2.6$-$3.2\sigma$.
Read the full abstract

High Energy Physics (Phenomenology)

Atmospheric neutrino up-scattering explanation of LZ 2026 excess

The recent observation of an isolated nuclear recoil at $248\pm 23\pm 23$ keV energy by the LUX-ZEPLIN (LZ) experiment has motivated the community to look for a new physics explanation, as the Standard Model background estimation fails to accommodate it. Most of the existing literature hitherto considers a galactic halo dark matter with a heavier partner. In this work, we traverse the alternate route of atmospheric neutrino ($\nu$) up-scattering, thus producing a massive beyond standard model (BSM) particle $\chi$ with mass $m_\chi$. The kinematic requirement of such a scattering demands a higher energy of the incoming neutrinos than elastic scattering, explaining the absence of excess events from solar $\nu$ scattering. However, for high energy neutrinos in the tail of the atmospheric neutrino flux, the kinematic threshold no longer prevents events at low recoil energies. To avoid an excess of such events, we consider a momentum dependent operator, which suppresses the event rate at low recoil energies and allows a single event in the high recoil energy window. Such up-scattering on nuclei ($\mathcal{N}$), $\nu \mathcal{N}\to \chi \mathcal{N}$, can be naturally realized in sterile neutrino models, though we keep our analysis generic without specifying $\chi$. We identify the region of parameter space that can produce such an isolated event, assuming a pseudoscalar mediator with mass $m_\phi$ that couples to the $\nu$-$\chi$ pair with strength $y_\chi$ and to up quarks with strength $y_u$. For example, $m_\chi \simeq 1.13$~GeV, $m_\phi \simeq 494$~MeV and $\sqrt{y_\chi y_u} \simeq 0.33$ can account for the observed event while remaining allowed by other existing constraints.
Read the full abstract

Human-Computer Interaction

Toward Frontier-Quality Declarative UI Generation at Small-Model Cost

Declarative UI protocols such as A2UI let applications generate interactive UIs by selecting pre-built components from a catalog and binding their props to application data, rather than emitting frontend code from scratch. This contract is attractive for production systems because of safety and consistency. An open question is: can low-latency and low-cost small models achieve the required quality for A2UI-based UI generation? To answer this, we systematically study three controllable design choices for catalog-conditioned A2UI generation: supervised fine-tuning (SFT) data construction method, model size, and component-catalog size. Across two React/TypeScript domains and four base checkpoints spanning two model families (Qwen 3.5 0.8B/2B/4B; SmolLM 3B), we find: (i) a 4B fine-tuned student recovers ~98% of teacher semantic quality and ~97% of teacher visual quality at more than an order of magnitude lower cost than frontier API calls; (ii) both augmented strategies (Perturbed-catalog and Constrained-GT) Pareto-dominate the unaugmented Full-catalog baseline, while specializing on different axes; (iii) even small models can handle and benefit from relatively large component catalog size. We distill these results into practitioner-facing trade-offs and deployment recommendations across the three design choices.
Read the full abstract

All authors Yingxiang Yang, Weihang Xiao, Ben Bullough, Tushar Deshpande, Niresh Agarwal

Computer Vision

Seeing Before Synthesizing: VLM-Guided Transition Event Discovery for Weakly-Supervised Dense Video Captioning

Weakly-Supervised Dense Video Captioning aims to localize and describe multiple events in untrimmed videos given only an ordered set of event-level captions per video. Recent work synthesizes auxiliary transition captions via LLM to provide additional vision-language alignment, but these captions lack visual grounding and are rigidly assigned to every inter-event gap at a fixed location and duration. To address these, we propose Seeing Before Synthesizing (SBS), a framework that adaptively provides visually grounded linguistic guidance only where warranted. Leveraging a VLM, we generate frame-level narratives for the inter-event gaps and detect transitions from the semantic variation across them. For identified transitions, we then refine inter-event temporal masks by blending the temporal midpoint with the semantic change point and selecting the width that maximizes vision-language alignment. Experiments on ActivityNet Captions and YouCook2 demonstrate state-of-the-art performance in both captioning and localization.
Read the full abstract

All authors Ye-Chan Kim, Seunghee Choi, SeungJu Cha, Si-Woo Kim, Hwiseon Kim, Hyungee Kim, Dong-Jin Kim

Quantum Physics

Quantum communication and Bell nonlocality require infinite classical communication to simulate

A quantum system of any fixed dimension can be prepared in a continuum of states, yet it cannot be used to transmit an unlimited amount of classical information. Similarly, the correlations observed between measurement outcomes on separate parts of a shared quantum system can be stronger than classical correlations, but they cannot transmit information. These fundamental limitations suggest that the statistics observed from quantum communication and quantum correlations may admit a simulation using a finite amount of classical communication. This expectation is confirmed in the smallest nontrivial quantum dimension, with two classical bits being necessary and sufficient to exactly simulate qubit communication and all correlations between qubits. Despite significant efforts during the previous decades, this remained the only solved case. Here we resolve both problems for every quantum dimension. The solution reveals an unexpected qualitative transition starting at dimension four: no finite amount of classical communication can exactly simulate ququart communication nor all quantum correlations of two entangled ququarts, even with unlimited shared randomness. One might have expected this transition, if it existed, to appear already for qutrits. Instead, we construct an explicit protocol that exactly simulates qutrit communication using $357$ classical bits, and consequently, all correlations of two entangled qutrits.
Read the full abstract

High Energy Physics (Phenomenology)

Seasonal dark matter from the LUX-ZEPLIN high-energy event

The LUX-ZEPLIN experiment has reported a single nuclear recoil at $248\pm23({\rm stat})\pm23({\rm sys})\keV$ in an extended search window. Inelastic dark matter, in which the scattering is endothermic, is a natural mechanism for producing a signal at such high energy. We fit the event to the dark matter mass and the mass splitting between two dark matter states, and find that the data favour a splitting pushed close to the kinematic limit set by the escape speed of the halo. The signal is then supplied by dark matter in the tail of the halo speed distribution. This drives a strongly seasonal signal: annual modulation rises above $50\%$ across the favoured region of parameter space and even reaches $100\%$ for some parameters, implying the signal vanishes for part of the year. Such a strong modulation distinguishes this interpretation sharply from a non-modulating signal, and we quantify the number of events needed to establish it. Experiments with larger target masses than LZ's, notably XLZD and PandaX, will be able to test this directly.
Read the full abstract

Computation and Language

Knowledge Acquisition During Pre-training? Large Language Models Learn Better With Auxiliary Views

Gaps remain in our understanding of how large language models (LLMs) acquire knowledge during pre-training. We find that auxiliary views, conceptual reformulations, are key to this process. We design controlled experiments to isolate this. First, holding the token budget fixed, allocating tokens from paraphrases to auxiliary views improves learning even for factual recall, which we use near-verbatim sentences from the original document to test. Second, the effectiveness of auxiliary views is not contingent on the strength of the model that generates them. Third, we establish that paraphrasing is helpful but only at smaller batch sizes. Fourth, we identify types of knowledge that aid learning in the presence of prior knowledge gaps. Finally, we examine how these effects manifest mechanistically via layer-wise biases and compression. Together, our findings suggest that auxiliary representations of knowledge are a key factor in the success of pre-training, and explains why data diversity matters. Further, our results provide clear recipes for synthetic data augmentation along with implications on the self-improvement in LLMs.
Read the full abstract

All authors Joseph Lee, Yidi Huang, Dokyoon Kim, Shu Yang, Li Shen

Representation Theory

Standard Representations of Crystallographic Coxeter Groups

This dissertation studies Cartan representations: representations of certain Coxeter groups as the Weyl groups of Kac-Moody algebras. It sets out and proves conditions for these representations to be isomorphic to each other, the reflection representation of the Coxeter group, and their dual representations, and gives partial results for the same over fields of finite characteristic, as well as clarifying what it means for a Coxeter group to be crystallographic and giving conditions for the Weyl group of a Kac-Moody algebra to be such. Some illustrative examples are also provided.
Read the full abstract

Quantum Physics

Measurements on the separated subsystems of an entangled state

The entangled states of composite quantum systems are well studied. The particle-like nature of these systems also means that, while entangled, they can be physically separated and measurements performed on the separated subsystems. Measurements on the separated subsystems A and B should pertain to vectors in the local Hilbert space of each subsystem, but to date it has not been clear how to elucidate the relevant states of the separated subsystems because it is not obvious how to resolve them from states given in the tensor product basis, except for the separable states. Here it is shown that the projections of any general entangled state that are detected by measurements on the separated subsystems can be obtained considering the corresponding (cosets of) states in the free vector space from which the tensor product space is defined. The result eliminates the need to invoke random collapse, and from this perspective nonlocality arises because of the way measurements on each separated subsystem projects possible measurement outcomes.
Read the full abstract

Artificial Intelligence

A Computationally Feasible Framework for Causal Probabilistic Explanation

Explaining why a specific outcome occurred, and which inputs deserve the blame or credit, is central to philosophical, scientific, and policy analysis. Existing tools split into two camps. The theory of actual causality (AC) gives principled verdicts, but only for toy-sized models, because computing them requires enumerating counterfactual scenarios. Scalable attribution methods like SHAP (or even causal SHAP) at least partially ignore the causal structure that generated the data, and can give answers that conflict with a careful causal analysis. We close this gap with Probabilistic Causal Impact (PCI). PCI builds on actual causality and on Pearl's notions of probability of necessity and sufficiency, but recasts the question of explainability as an estimation problem on a probabilistic causal model that is easily approximated via Monte Carlo. By specifying a distribution over "candidate explanations," a distribution over counterfactual values, and a scoring function, PCI provides tractable, causally grounded, graded explanations, generalizing AC and Pearl's probability of causation as degenerate cases. We evaluate PCI in synthetic and real-world examples, spanning consistency checks with AC, scaling experiments, complex continuous-valued dynamical systems, and a real-world deployed causal machine learning model trained on millions of datapoints.
Read the full abstract

All authors Rafal Urbaniak, Sam Witty, Daniel Waxman, Andy Zane, Poorva Garg, Emily Bunnapradist, Sankaran Vaidyanathan, Jack Feser, Drew Lehe, Eli Bingham

General Mathematics

Catalan's constant is irrational

Whether the constant $$G=\sum_{k=0}^\infty\frac{(-1)^k}{(2k+1)^2}=\frac1{1^2}-\frac1{3^2}+\frac1{5^2}-\frac1{7^2}+\cdots$$ introduced by Catalan in the nineteen century is irrational, is a long-standing open problem. In this paper we prove the irrationality of $G$ via using suitable weights.
Read the full abstract

High Energy Physics (Phenomenology)

Confronting the Higgsino Interpretation of the LZ Event with the High-Energy Sideband

The LZ experiment has reported a single 248 keV nuclear recoil event in their $2.84$ ton-yr exposure on a xenon target. The absence of lower nuclear recoil events is naturally explained by inelastically scattering dark matter (DM), for which the nuclear recoil energy spectra are shifted upwards from zero. Excitingly, the observed rate and spectra are consistent with a thermal $1.1$ TeV higgsino that explains all of DM with a mass splitting $\delta \sim 370 - 490$ keV between the two lowest-lying Majorana states, with the range depending on the underlying DM velocity distribution. In this work, however, we show that higgsino DM with these properties would generically produce more higher-energy recoil events in a high-energy sideband along the LZ $S1c$ energy axis, while the LZ experiment reports no events in that bin. This suggests a possible tension which could be further explored by the LZ Collaboration, as the acceptance in that bin is not public. We further show that a non-thermal $\sim$500 GeV higgsino may evade the sideband constraint and that future experiments using heavy targets can potentially test the higgsino interpretation of the event either way.
Read the full abstract

All authors Nicholas L. Rodd, Benjamin R. Safdi, Tracy R. Slatyer, Weishuang Linda Xu

Computer Vision

Zero-Shot Novel Depth Synthesis Using 3D Foundation Models Scene Representations

3D Foundation Models (3DFMs) such as VGGT have recently pushed the boundaries of 3D vision by predicting rich unified representations with feed-foward transformers. The scene representations learned by these models enable strong performance on multiple 3D vision tasks. In this paper, we investigate using their internal representations to infer 3D in the scene from new views. Our hypothesis is that in order to solve the task of 3D reconstruction, these models need to learn a representation that includes a large amount of general knowledge about 3D scenes. After showing that it is possible to decode hidden surfaces from internal 3DFM representations, we propose a method, Z3D, that estimates pointmaps in unseen views by doing latent diffusion on 3DFM representation. We show that Z3D can predict realistic depth maps for new views across multiple datasets.
Read the full abstract

Computation and Language

Last Translation Benchmark

For scientific progress, we need benchmarks that test the limits of state-of-the-art models, and evaluation methods that inform us about failure cases. As models get stronger, standard benchmarks for machine translation are approaching saturation. Further, automatic translation metrics are unreliable, opaque, and vulnerable to reward-hacking. Even gold human evaluation is not problem-free, because it often lacks reproducibility, objectivity, and scalability. Overall, this prevents us from tracking progress in the field and identifying pathways for improvement. We introduce the Last Translation Benchmark, a collection of human-authored and peer-reviewed examples (texts, images, audio, videos) that break leading machine translation models. We also present a new evaluation approach: each example comes with handcrafted verification rules describing concrete failure cases on that example, therefore allowing reliable and actionable future evaluation. The Last Translation Benchmark is a live dataset that accepts ongoing contributions. The latest version is LTBv1, containing accepted contributions prior to September 1st 2026, with future releases planned as new data is continuously collected.
Read the full abstract

All authors Vilém Zouhar, Niyati Bafna, Mukund Choudhary, Maike Züfle, Sara Rajaee, Pinzhen Chen, Jannis Vamvas, Sara Papi, Ona de Gibert, Bhavitvya Malik, Eliya Habba, Orfeas Menis Mastromichalakis, Patrícia Schmidtová, Michelle Wastl, Sheriff Issaka, Leshem Choshen, Stella Biderman, Antonis Anastasopoulos, Jan Niehues, Rico Sennrich, Mrinmaya Sachan, Ondřej Bojar, Kenton Murray, Jörg Tiedemann, Alham Fikri Aji, Philipp Koehn, Christof Monz, Alexandra Birch, Sowmya Vajjala, Chalamalasetti Kranti, Cristina España-Bonet, Nobin Sarwar, David Kaczér, Sourajit Saha, Jonathan Tonglet, Shunta Asano, Malik Marmonier, Daban Q. Jaff, Vaisakhi Mishra, Hend Al-Khalifa, Gabriele Sarti, Nils Rehlinger, Juan Daniel Cuervo Villa, Dominik Macháček, Saugata Purkayastha, Jagannathan Ramanujam, Shubhashis Roy Dipta, Aviral Nigam, Shuaib Shuaib Yusuf, Heejin Do, Jonathan Yahav, Johannes-Rudolf David, Maria Carmen Staiano, Zuzana Nadova, Fred Philippy, Ron Keinan, Maria Lymperaiou, Silvia Casola, Fabian Retkowski, Andrés Jerez, Hanna Yukhy- menko, Sangwon Ryu, Avantica Vempati, Sukannya Purkayastha, Adrian Cosma, Erivan Inan, Vitalii Babenko, Wafa Aissa, Valentin Scourneau, Fatima Haouari, Venkata Prasanth Kumar Gummadi, Mehdi Jafarzadeh, Manon Reusens, Kaiser Sun, Lukas Edman, Shaomu Tan, Giuseppe Gallipoli, Pawan Sasanka Ammanamanchi, Manar Ali, Mohammad Sadegh Gholizadeh, Dipankar Srirag, Marek Šuppa, Javier García Gilabert, Ruta Binkyte, Ana-Maria Bucur, Sabry E. Farrag, Youssef Saber, Yihong Liu, Theresia Veronika Rampisela, Christian Hoang, Nicoleta Cojocaru, Jan Kocoń, Jean Maillard, Xiaochuang Yuan, Sina Ahmadi, Daryna Dementieva, Philipp Mondorf, Kaustubh Dhole, Valmik Nahata, Roman Wixinger, Amir Hossein Yari, Shenbin Qian, Manuel Tuor, Fida Mohammad Thoker, Sergey Troshin, Lance Calvin Lim Gamboa, Amir Arsalan Rezapour, Kätriin Kukk, Koel Dutta Chowdhury, Shaswati Saha, Seth Aycock, Bo Chen, Linh Vu, Vatsal Venkatkrishna, Shayan Bali, Arafat Ahsan, Luan Thanh Nguyen, Hassan Soliman, Ngoc Quynh Tram Do, Azmine Toushik Wasi, R. Damanhuri, Marius Huber, Kazuki Egashira, Jimson Paulo Layacan, David Africa, Vladislav Poritski, Mike Zhang, Deep Shah, Abdulaziz Nura Kani, Luis Frentzen Salim, Paul Gavrikov, Bello Umar Bello, Ayush Sunil Munot, Anumit Garg, Yolanda Xavier, Qiaoyuan Zheng, Kawsar Ahmed, Debanshu Das, Zimu Wang, Gengyu Rao, Kamile Dementaviciute, Farhan Farsi, L D M S Sai Teja, Dawei Zhu, Yi Fan, Wei Liu, Marco Gaido, Elias Herranen, Sankalan Pal Chowdhury, Karen Sanchez, Guy Kaplan, Farzad Shami, Ashok Urlana, Amir Hossein Kargaran, Sofie Goethals, Priyaranjan Pattnayak, Oksana Volchek, Marii Ojastu, Hongbin Na, Emilian Radoi, Chenyi Zhao, Carlos Hinojosa, Andrei Niculae, Andrea Gregor de Varda, Zaid Alyafeai, Tomasz Limisiewicz, Reem Alzahrani, Pouya Sadeghi, Nehal Kathrotia, Mateusz Lango, Enzo Doyen, Alex Flückiger, Ulysses Sekai Tully Carr, Samuel Simko, Ritwik Tiwari, Rishit Dagli, Isaac R Caswell, Bowen Yi, Aicha Chorana, Selja Keränen, Sadiksha Chitrakar, Muhammad Ravi Shulthan Habibi, Joy Olusanya, Bishal Shrestha, Zhengxiang Wang, Vivek Harsha Lakkamaneni, Sophia Conrad, Panayiotis Panayiotou, Nazia Tasnim, Marta Punsola Munárriz, Marko Culjak, Luis Lara, Jenny Chim, Jannatul Nayem, Fidel Rodríguez Velásquez, Eran Yahav, Blanka Kövér, Beatrice Savoldi, Anmol Goel, Aishik Mandal, Tosin Adewumi, Raoyuan Zhao, Mykola Haltiuk, Antonia Karamolegkou, Yuxing Lu, Thura Aung, Naser Almousa, Tommaso Cerruti, Raia Abu Ahmad, Beni Egressy, Alireza Pakniat, Stéphane J.P.S. Thunus, Rachel Bawden, Lena Libon, Samridhya Biswas, Prakhar Gupta, Nusrat Jahan Lia, Nguyen Tai, Natchapon Jongwiriyanurak, Minh Ngoc Do, Ivan Barać, Dzmitry Kuzmin, Badal Nyalang, Antoine Taroni, Andy Catruna, Rushikesh Zawar, Roland Aydin, Pavel Stepachev, Ilai Yaron Levy, Andreas Simons, Rayyan Merchant, Ziyi Yang, Samuel Frontull, Kenneth Enevoldsen, Harris Abdul Majid, Tim Graf, Tatiana Bielakova, Sharifa Djurabaeva, Shaoxiong Ji, Jirui Qi, Ayla Rigouts Terryn, Yurii Paniv, Xiyan Fu, Sunisth Kumar, Shree Harsha Bokkahalli Satish, Papa Abdou Karim Karou Diallo, Mengyu Ye, Maximilian Vieweg, Matija Akrap, Kristýna Onderková, Joseph Attieh, Ivan Yuri De Leon, Ibrahim Baroud, Esrael Teferi Tensay, Elisabeth Fittschen, David Dukić, Benoît Sagot, Jingwei Ni, Yu Fan, Juri Opitz

Artificial Intelligence

Rethinking On-Policy Distillation of Large Language Models II: One Training Example

On-policy distillation (OPD) combines student-generated rollouts with dense token-level supervision from a teacher. Existing work has mainly studied its algorithmic behavior, leaving the role of training data unclear. We examine this role at the data-minimal limit by training on a single query. One-shot OPD keeps improving for hundreds of steps and recovers most of full-data OPD's gain across task domains and model families. We explain this result through the states visited during training and the rate at which the student aligns with the teacher. We measure state coverage, the fraction of the states full-data OPD visits that a query set's rollouts reach. A single query already reaches \(71.5\%\), most of it within the first 100 steps. Adding semantically distinct queries raises coverage and validation accuracy together, until 16 queries reach \(98.9\%\) and match full-data training. Yet alignment slows at a similar pace whether OPD trains on one query or the whole dataset, and even a fixed set of states takes hundreds of steps to absorb. OPD is therefore data-overfed but algorithm-starved. Its rollouts quickly expose broad supervision, while the student absorbs that supervision increasingly slowly. The state-coverage result extends to multi-teacher OPD, where 16 semantically diverse queries per domain match full-data MOPD. As a further stress test, content-light templates and off-domain WildChat queries also approach the real-query baseline. Task content and induced state coverage can therefore come apart. We hope these findings direct future work toward the step efficiency of OPD, and prompt a re-examination of the data and the mechanisms behind its recent successes in frontier post-training.
Read the full abstract

All authors Zixuan Fu, Bingxiang He, Yuxin Zuo, Haohuan Huang, Jinqian Zhang, Ruhang Xiao, Cheng Qian, Qinyu Luo, Huan-ang Gao, Yudong Wang, Zhiyuan Liu, Ning Ding, Chaojun Xiao

Optics

Twin-photon generation in a silicon nitride microresonator

Photonic chips with silicon nitride ($\mathrm{Si_3N_4}$) microring resonators are well established as heralded single-photon sources, but their operation as frequency-degenerate twin-photon sources has not previously been demonstrated. Here, we realise a twin-photon source at telecommunication wavelengths in a $\mathrm{Si_3N_4}$ ring microresonator via an inverse four-wave mixing (FWM) process, in which two photons from spectrally distinct pumps are converted into a pair of identical twin photons. The measurements show a maximum coincidence-to-accidental ratio (CAR) of $5.4\pm0.6$. In addition, the microresonator functions as a heralded single-photon source through pump-degenerate spontaneous four-wave mixing (SFWM), exhibiting a spectral purity of $P=0.67\pm0.05$ and a heralded anti-bunching of $g^{(2)}_h(0)=0.0042\pm0.0015$. Together, these results demonstrate both photon-generation schemes on a single integrated $\mathrm{Si_3N_4}$ platform, highlighting its potential for scalable, tailored quantum light generation.
Read the full abstract

All authors Franz Pacher, Haochen Yan, Alekhya Ghosh, Arghadeep Pal, Toby Bi, Hao Zhang, Lixing You, Hao Li, Daniela Salvoni, Shuangyou Zhang, Pascal Del'Haye