Science & Mathematics

\emph{Multiparty session types} (MPST) is a type discipline for ensuring distributed programs are implemented correctly \emph{by construction}. MPST starts by specifying a choreographic protocol as a \emph{global type} and \emph{projects} it onto each participant, generating \emph{local types} with a guarantee of type soundness and deadlock-freedom. A central issue of MPST is that its type soundness is fragile, as it must be proven each time with respect to a chosen projection algorithm under specific semantics with a specific program correctness goal. To overcome this problem, we propose general criteria for a \emph{good projection} -- a set of requirements for projection algorithms by which type soundness, deadlock-freedom and liveness of distributed programs are \emph{automatically} derived. Using the criteria, we prove that 8 distinct projections in the literature are good under both synchronous and asynchronous semantics, and that 4 are incorrect, failing type soundness. We extend the good projection criteria to mixed choice global types and asynchronous communication optimisations, proposing new ``good projection'' algorithms. We apply properties of good projections to prove \emph{separation} and \emph{expressiveness} results over the projection algorithms and MPST typing systems. Having a generalised typing system that is robust with respect to the good projection criteria, we provide an implementation with mixed choice global types that is \emph{modular} with respect to a chosen projection algorithm and desired program properties. Our separation results help programmers make informed trade-offs between complexity and expressiveness when choosing projection algorithms, and our theory ensures their chosen algorithms are correctly applied.

A Berry-curvature dipole can produce a nonlinear Hall current even when the linear Hall response vanishes, but symmetry alone does not determine its magnitude or magnetic tunability. The interplay of structural symmetry, moment orientation, and carrier occupation is examined in a honeycomb model. Curvature maps and the band-summed dipole integrand reveal how velocity and the occupation window near the Fermi level select signed contributions. Removing threefold rotation permits a nonmagnetic dipole perpendicular to the surviving mirror. Generic magnetic reorientation activates the second component and tunes both components through positive and negative values. Global reversal preserves the dipole while reversing the intrinsic linear Hall conductivity, allowing a finite bulk dipole contribution from equally populated reversed domains under uniform-field averaging.A symmetry-constrained polynomial model predicts both dipole components for 256 held-out spin configurations, achieving coefficients of determination $R^2 > 0.9995$ at the reference chemical potential. At fixed carrier density, orientation-dependent shifts of the chemical potential modify the response and yield slightly larger prediction errors for the chosen model. Across the chemical-potential sweep,the prediction accuracy decreases more strongly for $D_y$,revealing an energy dependence of the surrogate's performance. Within semiclassical transport with a constant relaxation time, the calculated dipole determines the angular dependence of the quadratic current and its rectified and second-harmonic components. These results provide a quantitative representation of magnetic-orientation-dependent Berry-curvature dipoles and show how carrier occupation influences both the response and its predictability.

Collaborative perception enables agents to retrieve the scene semantics and geometry beyond the field of view of any individual agent. This broader scene understanding enhances transportation safety and improves efficiency in multi-robot systems. However, existing collaborative semantic occupancy methods rely on dense 3D voxel annotations and are limited to recognizing objects within a predefined set of semantic classes. To overcome these limitations, we present CoGaussian, an open-vocabulary Gaussian-based collaborative semantic occupancy prediction framework. CoGaussian represents a scene as a set of foundation model-aligned sparse 3D Gaussians, which serve as a compact representation for communicating and fusing geometric and semantic information across multiple agents. On the Semantic-OPV2V benchmark, our method outperforms existing open-vocabulary single-agent approaches, highlighting the effectiveness of open-vocabulary collaborative scene understanding.

Process mining typically explains an event log with a single model, clustering, or classifier fitted to all cases, although real logs often mix cases that follow different sub-processes. Trace clustering separates sub-populations, but it fixes one notion of trace similarity for the whole analysis. We propose TracePrism, which treats process discovery as iterative partitioning with three characteristics. The trace encoding is adaptive: it is chosen for each working set of remaining traces from a library of encodings, with suggestions scored for that set. The partitioning is iterative: the analyst forms a candidate group based on a rule, a dominant topic, a cluster over topic weights, or a region of an embedding, accepts it if it is coherent in both process structure and outcome, and continues with the remaining traces. Decisions are revisable and recorded: a group can be reopened under a different working set or encoding, and the findings repository keeps both the earlier decision and its revision. We implement TracePrism as an open-source prototype and demonstrate it in usage scenarios on three real-world event logs. The scenarios also revealed behavior that we had not anticipated, such as process drift.

This chapter introduces data ethics for translation and discusses a number of ethical issues regarding data. These include issues of copyright, ownership and bias, which apply to all types of data, whereas there are further issues particular to minoritised language communities and those on either side of the digital divide, as data is implicated in power imbalances and dynamics. We close with a discussion of responsibility and regulation, and reflect on the growing attention on data ethics from translation studies researchers.

Envy-free cake cutting asks how to divide a resource so that no agent prefers another agent's share. Finding such a division with few queries about preferences has been a longstanding challenge. Aziz and Mackenzie gave the first query bound depending only on the number $n$ of agents: a tower of six $n$'s. Subsequent improvements culminated in Ye and Bai's single-exponential bound of $n^{O(1)}2^n$. A large gap remains between this deterministic upper bound and Procaccia's general lower bound of $Ω(n^2)$. Whether polynomially many queries suffice is a central open question in fair division. We show that a polynomial expected query bound is possible. Our randomised protocol uses only independent, uniformly sampled points from the cake as its source of randomness. For every profile of $n$ agents with nonnegative, additive, nonatomic valuations, the protocol has the following guarantees: (1) It terminates with probability one. (2) Every allocation it returns is complete and exactly envy-free. (3) Its expected number of Robertson-Webb queries is $\widetilde{O}(n^{30})$. The expectation is taken solely over the protocol's random choices, with no distributional assumptions on preferences.

We present Qwen-Audio-3.1-ASR, a family of audio-language models that builds upon its predecesor Qwen-Audio-3.0-ASR with improved speech recognition and expanded capabilities in speaker-diarized transcription and general audio understanding. The family retains support for 30 languages and 16 Chinese dialects, together with instruction-controlled capabilities including historical-context conditioning, hierarchical hotword customization, and native transcription polishing. Speaker-diarized ASR produces structured outputs containing speaker labels, timestamps, and recognized text, maintaining consistent speaker labels across turns and the representation of overlapping speech. The general audio understanding capabilities cover audio-, speech-, and music-captioning, temporal grounding and sound event detection, as well as audio question answering and multiple-choice reasoning. These tasks broaden the application scope from recognizing spoken words to interpreting speakers, acoustic events, and their temporal relations. To accommodate different deployment requirements, the family includes variants for asynchronous file transcription, low-latency streaming, interactive voice input, and non-streaming recognition, with speaker-diarized transcription and general audio understanding available in certain variants. Evaluations span conventional ASR, multilingual and dialectal speech, speaker-diarized transcription, and general audio understanding, demonstrating the capabilities of the Qwen-Audio-3.1-ASR family across transcription and broader audio analysis tasks.

Post-training quantization (PTQ) compresses the weights of a trained model with limited compute and data. Down to 4 bits it is near-lossless, but at 2 bits scalar PTQ configurations fall to chance-level accuracy. We study three choices usually treated as implementation details that matter little at 4 bits but decide 2-bit quality: the quantization representation, the calibration data, and the recovery fine-tuning data, across six dense and mixture-of experts models (ranging from $3.8$ to $32$B parameters). Representation is the largest factor and vector codebooks restore usable models where scalar PTQ pipelines collapse. Most of the remaining loss in accuracy is on \emph{generative} tasks, and the amount of performance drop is model-dependent. Furthermore, at a short recovery budget, calibration and recovery data are coupled. The best recovery data depends on the calibration data, so the two must be chosen jointly. The resulting recipe involves a vector codebook representation, a small task-aligned calibration mixture, and an inexpensive codebook-only recovery stage. Including codebooks, it uses about $2.3$ effective bits per weight and retains $89$ to $97\%$ of full-precision question-answering (QA) accuracy where scalar PTQ is at chance, and, including recovery, uses $1.4$ to $30\times$ less compute than the strongest quantization-aware-training (QAT) baseline.

How can we effectively oversee AI systems that surpass human intelligence? AI Safety via Debate proposes that two AI systems argue opposing positions and a judge, either a human or a weaker model, decides which is correct. By facilitating supervisors in providing high-quality training signals, debate aims to encourage truthful and safe behavior from AI systems whose capabilities exceed our own. However, human judgment is not neutral: people often rely on biases which advanced AI systems may learn to exploit. Drawing on insights from collective intelligence and social cognition, this paper investigates whether debate can guide human and LLM supervisors toward the truth despite their prior beliefs. We evaluate several protocol configurations. We also explore setups inspired by the wisdom of the crowd such as the number and diversity of judges, as well as interaction mechanisms including deliberation and voting. Overall, we find that debate increases accuracy relative to prior beliefs for human judges, whereas the consultancy baseline, in which a single expert argues for one answer, does not. For weaker language models, however, the results are mixed: debate provides no comparable gains in most settings and can even reduce accuracy. Among multi-judge protocols, weighted voting produces the highest accuracy for human supervisors, whereas for LLMs none of them significantly improves group accuracy, even as the number of judges increases. Notably, the behavior of minority judges in non-diverse deliberation protocols suggests that, under symmetric settings, incorrect positions could be more persuasive than correct ones. These findings support debate as a viable path toward scalable human oversight, but cast doubt on its benefits for weaker AI supervisors, raising concerns about fully automated training and control pipelines in high-stakes settings.

Autoregressive language models are trained by next-token cross-entropy, yet several recent systems also iterate a hidden state without emitting a token: continuous thoughts, pause steps, and recurrent depth. This note separates three risks on one latent process. The token risk $R_{\mathrm{tok}}$ scores a readout. The teacher-forced latent risk $R_1$ scores a one-step map that sees the true hidden state. The open-loop risk $R_H$ scores the same map after it has been iterated on its own outputs. If the Bayes next-token posterior is a function of a $k$-dimensional statistic with full-rank covariance, a width-$k$ latent, a linear least-squares transition, and a refit unembedding match the Bayes token predictor. The statistic is identified only up to an invertible gauge: whitening the residual stream does not change the token law once the unembedding is refit. The matching obstructions are sharp. A constant latent drives $R_1$ to zero while token cross-entropy falls back to the marginal entropy. A perfect token predictor can still hide a null-space walk whose open-loop error grows linearly in the horizon. A small next-latent loss does not certificate next-token prediction, and a small next-token loss does not certificate free latent rollout, unless a rank or injectivity condition is stated with the number.

Discrete diffusion models iteratively denoise a noisy sequence of categorical data such as text, enabling any-order generation with self-correction and in-filling capabilities. This makes them promising candidates for reasoning, planning, and constrained-generation tasks like coding and math. Yet masked and uniform diffusion erase a token's identity in a single corruption event, so their noisy states cannot explicitly represent partially resolved uncertainty in which several alternatives remain plausible. We close this gap with Binary Discrete Diffusion, a family of continuous-time discrete-diffusion processes whose noisy state is a set of candidate tokens per position, represented as a binary membership vector. Binary diffusion changes candidate sets through additions and removals, allowing information to be lost and recovered gradually while the diffusion state itself represents partial uncertainty. On Sudoku-Extreme, Boxoban, and graph coloring, the symmetric formulation of binary diffusion consistently outperforms strong, well-tuned masked and uniform baselines under matched architectures and controlled training, sampling, and tuning budgets, with its largest gains on the hardest instances, while remaining competitive on language modeling (Text8, TinyStories, OpenWebText). These results support candidate sets as an effective intermediate representation for iterative reasoning and show that the corruption kernel can materially change the computation learned by the reverse model.

Decentralised federated learning (DFL) trains personalised models on devices that average with their neighbours rather than with a central server. Real devices, however, are often offline for hours or days, and a client returning from an outage can recover only by aggregating the models its neighbourhood holds. We observe that the stalest of these models is usually the client's own, a case that existing staleness-aware rules, which discount only the models a client receives, do not address. We formalise aggregation after an outage as estimating the state of a fully available network and derive three principles. First, the receiver must be discounted like any neighbour. Every rule that weights it at least as much as each neighbour has an error that grows without bound with the outage, whereas discounting all candidates by relative staleness keeps the error bounded. Second, staleness must be measured as inherited lag rather than age, since local training resets a model's age but not the lag it absorbed from stale neighbours. We introduce the effective lag, which tracks this quantity at the cost of one scalar per message. Third, the discount must be steep, because absorbed lag persists in every later round. These principles define Self-inclusive Freshness-weighted Gossip (SFG), a family of drop-in aggregation rules with the communication and computation of decentralised federated averaging, whose guarantees hold for the stale receiver itself. On three image benchmarks and eight outage processes, including real smartphone availability traces, the lag-based members SFG-L and SFG-Cut consistently outperform decentralised federated averaging and recent age-of-information rules, and ablations show that each principle is necessary. The analysis also explains why tuned limited merging from gossip learning is effective.

Semantic data analytics systems integrate AI-powered operators into query pipelines, enabling declarative queries over structured and unstructured data. However, existing systems rely on per-tuple model invocations, ad hoc operator designs, and lack principled cost models, resulting in poor scalability and limited optimization. We present NousDB, a high-performance semantic analytics engine built on Apache DataFusion[22]. Through NousDB we propose an extended relational algebra with three semantic operators and formal algebraic properties that enable principled query rewriting and cross-operator optimization. We further propose a token-based cost model using equi-depth token histograms to optimize batch sizing under context-window constraints and provide statistical guarantees for safe batching. Experiments show that NousDB achieves up to x86 speedups with both local and enterprise LLMs, while leading competitors in token efficiency and accuracy. These gains extend to complex JOB and Stack workloads augmented with semantic operators, supporting both traditional and semantic analytics.

Despite the success of video world models, extending them to multiple agents is difficult because it requires generating several agent-specific video streams that are consistent in time and between agents. To address this gap, we introduce OverSeer, a multi-agent world model that augments egocentric agent views with auxiliary camera views. We suggest that, by fixing the auxiliary views in space, they act as a proxy for a world-centric representation: Unlike the agents' views, which can change entirely due to a simple change in agent position, changes in the auxiliary views directly map to localized changes in the world state. We show that these fixed views substantially help the model to maintain consistency. We also show that, due to their redundant and semi-static nature, the auxiliary views can be efficiently compressed with a delta-VAE that encodes only the changes relative to a static reference image, reducing the number of tokens required to process them. Experiments in Minecraft and Doom demonstrate improved world and agent consistency, with qualitative results on real-world multi-camera captures indicating that the approach is not tied to game engines. Our compression analysis further shows that delta-VAE substantially reduces the computational overhead of auxiliary-view generation. We conclude that these fixed views provide a simple and efficient way to improve multi-agent world models.

Statistical learning in high dimensions suffers from the curse of dimensionality. Fortunately, many functions of practical interest are compositionally sparse, meaning they admit a composition of local functions, each depending on only a few variables, allowing sample complexity to depend on local arity rather than ambient dimension. Yet realizing this benefit requires exploiting hidden compositional structure, which theory does not establish for practical gradient-based training. We study this gap with CSFRBench, a controlled benchmark that hides the target's compositional computation during training but retains it for diagnosis, and find that dense networks and sparsity-based baselines leave much of that benefit unrealized. Inspired by the spatial organization of biological neural circuits, we introduce Learned Wiring (LeWi): each neuron has a learned low-dimensional embedding, only nearby neurons connect, and embeddings and weights are trained jointly. LeWi improves sample efficiency on targets with hierarchical structure, intermediate reuse, or independent modules. Across 32 additional targets generated by independently sampled computation graphs, LeWi is the most sample-efficient method not given the target graph, with 27% lower mean learning-curve area than dense networks. Dense networks perform better on targets with a single high-arity function. LeWi's representations and connectivity align with the target computation. On 17 world-model prediction tasks spanning control, navigation, and manipulation, LeWi beats the dense baseline on 16 and wins 99/119 baseline comparisons. These results show that jointly learning connectivity and weights can exploit computational structure to improve sample efficiency.