Papers

Liu introduced generalized Fibonacci gates on domains of size 3 and 4, proved that Holant problems built from them are tractable by a cut-and-merge algorithm, and conjectured that Fibonacci gates with matching polynomial-time algorithms exist on all higher domains. We extend the definition uniformly to every domain size $k$ and show that a family of domain-$k$ generalized Fibonacci gates is the same thing as a $k$-dimensional commutative Frobenius (equivalently, Artinian Gorenstein) algebra $A$ equipped with an orthonormal basis containing $1$. Under this dictionary Liu's parameter constraints are exactly associativity of $A$, a gate is an element $w\in A$, merging two gates is multiplication in $A$, and the Holant value of a connected graph $G$ is $\tau\bigl(h^{\,|E|-|V|+1}\prod_v w_v\bigr)$ for a fixed “handle” element $h\in A$. This proves Liu's conjecture for every $k$, with an algorithm whose cost is polynomial in $k$ as well as in the graph. We also give a closure criterion for linear recurrences of any depth (a balance condition implies closure, and is necessary when the unit is a combination of the colours), and we map the landscape: semisimple algebras are exactly Liu's orthogonal frames, smoothable algebras with nilpotents are collided frames, and families whose algebra is not smoothable (“frameless” families) exist exactly when $k\ge 14$. Finally we show that every non-semisimple part of a family is invisible to the Holant on connected graphs with two or more independent cycles. This note states the results and sketches the proofs; the proofs of the main theorems are short and are given essentially in full.

Wireless sensing systems often transmit highly redundant observations, making communication a bottleneck for low-power devices. We present Memento, a communication system that exploits recurring content by transmitting observations by reference. The sender embeds each observation and searches a shared vector table stored in persistent memory. When it finds a sufficiently similar prior observation, it sends a compact reference; otherwise, it sends the observation itself and adds it to the table. This design shifts work from wireless communication to local computation and storage, without requiring a fixed task-specific model. We implement Memento on a microcontroller with an SD card and evaluate it on three image datasets and one radar dataset. Memento provides favorable accuracy-communication and accuracy-energy tradeoffs against compression-based and task-oriented baselines. At high-accuracy operating points, it matches cloud classification accuracy on three workloads while using 5.5 to 27.1 times less energy, and comes within 0.14 percentage points on a fourth while using 14.7 times less energy. These results show that retrieval over persistent local memory can reduce communication costs in recurrent sensing workloads.

Frontier AI agents increasingly act in environments where seemingly solvable tasks may lack prerequisites, contain conflicting requirements, or reveal blockers only during execution. We formulate agentic self-awareness (ASA) as the alignment of task-grounded feasibility judgments (know), warranted behavior (act), and responses to reconsideration (hold), distinguishing what a task permits from what an agent can solve. We build paired feasible and blocked tasks in Terminal-Bench 2.1 and retail $\tau^2$-bench and introduce an authorship-aware M0-M3 rubric that separates recognizing a blocker from taking the certified deferral action for the correct reason. Separate probes assess stated feasibility beliefs and responses to a single challenge. In a Terminal-Bench protocol baseline, blocker recognition reaches 84.4%, but strict self-awareness (SA) reaches only 40.6%. Harness search without model-weight updates increases strict SA from 45.3% to 93.8% on reused Terminal-Bench search tasks; selection for abstention alone, however, can reduce success on feasible tasks. Independently initialized joint-objective searches improve both blocked-task SA and feasible-task solve on their own search sets. Retail held-out gains remain limited, and Terminal-Bench blocked-task transfer is not measured. This evaluation pipeline makes the recognition-action gap measurable and sets a two-sided target for improving agent decisions under changing task evidence.

Surface navigation on the Moon is subject to extreme illumination contrast and deployment-time perceptual disturbances. While modular navigation pipelines have long supported planetary rover navigation, robustness is compromised by measurement errors that propagate through hand-engineered components. End-to-end learned policies are an alternative approach to navigation, which maps observations directly to control actions without intermediate components. Vision-language-action models, in particular, have driven significant breakthroughs in robotics and autonomous vehicle research. By leveraging internet-scale vision-language pretraining, these models can be exploited for their commonsense reasoning and visual understanding to provide robust generalization with limited task-specific data. However, their applicability and performance have yet to be characterized for lunar surface operations. In this study, we design a vision-language-action policy for autonomous lunar rover docking and characterize its performance relative to a modular navigation pipeline. We develop a scalable training framework that generates safety-aware expert demonstrations in a high-fidelity lunar simulation environment. In closed-loop evaluations, our policy performs competitively relative to the modular baseline under perceptual disturbances unseen during training. To address the tradeoff between compute consumption and closed-loop safety, the policy adapts its execution horizon by determining how far the rover may safely travel in open loop. Adaptive-horizon execution reduces inference frequency while limiting hazard-contact rates relative to fixed horizons in the evaluated scenarios.

LLM-based auto-formalization systems can now generate formally verified Lean code at unprecedented rates. However, a formalization advances mathematical research only when further mathematical results can be built on it, especially when its definitions and helper lemmas are general enough to be imported and used idiomatically. We propose definition grounding, where definition formalizations are continuously judged and revised, as a method to ensure formalization quality. We find that agents formalizing definitions from textbook text often produce definitions that later statements cannot use, and only 25%-50% of their definition sets are compatible with all later statements. However, clarifying the informal definitions can improve this rate up to 77.5% averaged over all models. Furthermore, stating a small set of basic properties without proving them can improve model judges' capability to detect definitions with compatibility issues, at only about 11% of the definition formalization cost. These results are verified with ReuseBench, a benchmark we curated for formalizing new mathematical fields and theories.

Operational reporting for high-performance computing (HPC) facilities requires combining information from monitoring services, schedulers, incident and maintenance records, storage measurements, periodic reports, and operator investigations. Producing a reliable report requires more than retrieving relevant records: the evidence must refer to the correct system, metric definition, time period, and source of authority. We present REPORT-OPS, a hybrid reporting agent in which an LLM interprets requests, plans evidence collection, evaluates possible explanations, and drafts reports, while author-seeded operational policies and automated checks determine whether the available evidence supports the requested conclusion. Depending on the evidence state, the system answers the request, reports a qualified conclusion, or withholds an unsupported claim. During offline refinement, an LLM critic proposes bounded policy changes that a human approves or rejects without editing; accepted candidates are promoted only after automated target, held-out, and regression testing. We evaluate REPORT-OPS on 68 sanitized reporting queries constructed from 50 base scenarios in a frozen HPC operational corpus, using three agent models and two LLM judges. REPORT-OPS achieves 68.63% Pass@80, compared with 7.02% for the strongest retrieval baseline. Evidence gating increases claim-support precision from 40.58% to 90.06%. On the 13-query explanatory subset, explanation accuracy reaches 55.54%. Across five rounds of human-supervised refinement, localized policy updates correct 91.17% of targeted failures and improve performance on related held-out cases by 9.33 percentage points, with no observed regression breaks under release gating. These results demonstrate improved evidence-grounded reporting and within-environment policy reuse; they do not establish unrestricted causal discovery or cross-facility generalization.

The Continuous Thought Machine (CTM) is a recurrent network that attends to inputs and makes predictions based on the synchronization between its neurons. On tasks spanning many domains, it displays emergent thought processes that have not been explained. We report a number of findings about how the CTM classifies images, focusing on ImageNet. Each thinking step is an attracting but strongly non-normal map exhibiting balanced amplification (Murphy and Miller, 2009); while attraction is already present at CTM initialization, training highly promotes non-normality. By Banach's theorem, the neural state has a locally unique attracting fixed point, which moves at every step, and the thought trajectory is the curve these fixed points trace, followed one tick behind. Most of that curve's motion is shared across images, and visible as a previously reported traveling wave of activity across neurons. The attention schedule along the thought is largely generic: shuffling queries across images at every tick costs only 7 percent of classification accuracy, while image-independent increments predict the next query with R^2 = 0.98 (held out). Our results suggest that the CTM has found a single strategy to process all ImageNet images.

Myopic vision degrades the quality of human vision due to spherical eye defocus aberrations that impair the perception of fine spatial details. To address this limitation, this work proposes a deep learning-based approach to computationally design transformed images that allow individuals with myopia in fixed distance to perceive scenes with enhanced clarity. The method trains a neural network to counteract defocus aberrations modeled using the second-order Zernike polynomial $Z_2^0$. The approach employs a U-Net architecture trained with a loss function combining reconstruction error minimization with spatial difference structural preservation through $\ell_1$ regularization. Training is carried out exclusively on the KITTI dataset, while testing is performed on DIV2K and ImageNet datasets, demonstrating generalization across diverse image distributions. Under myopic conditions modeled by $Z_2^0$, baseline performance shows visual degradation with PSNR values ranging from 19.47 to 27.42 dB across datasets and severity levels. The proposed gradient-regularized method achieves consistent improvements, delivering PSNR gains ranging from 1.48 to 7.34 dB and significantly lower perceptual compared to classical baselines.

Large language models are increasingly deployed as teams of agents that hold different private information about a shared task. How should such agents concisely communicate in a way that promotes aggregation of decision-relevant information? Communicating only proposed actions can discard decision-relevant uncertainty: two agents may prefer the same action for different reasons. We study an alternative interface in which agents communicate calibrated beliefs over actions. We first analyze a simple synthetic setting that captures the structure of multi-round reasoning under partial information. In this setting, we show theoretically and empirically that communicating unbiased probabilities can be strictly more powerful than communicating actions. We then test the same principle in a collaborative maze-solving task with trained transformers and pretrained LLM agents. Our results demonstrate that probability communication outperforms action communication when the exchanged predictions preserve calibrated uncertainty. When communicated probabilities are biased, we show that the benefits of probability communication can disappear. We give a post-hoc conversational calibration intervention that consistently improves decision-making by correcting these biases both in a toy transformer setting and when pretrained LLMs use the corrections in context. These results suggest that probabilistic communication is useful for multi-agent collaboration not merely because it transmits more information, but because calibrated uncertainty gives other agents a reliable object to update on.

Understanding how large language models (LLMs) evaluate gameplay experience is important for using their feedback to guide game design. However, overall ratings alone provide limited insight into which aspects of experience are associated with these judgments and how they can inform concrete game changes. We introduce PREFER, a framework that uses a shared set of experience dimensions adapted from established psychometric questionnaires to benchmark how models evaluate gameplay and to guide persona-specific game refinement. Through these dimensions, we characterize the perspectives models offer as virtual playtesters by examining how their overall assessments relate to specific aspects of experience and to the gameplay they produce. To examine how the same dimensions can guide refinement for players with different preferences, we use personas to simulate distinct player profiles. For each persona, PREFER uses experience-preference relationships to select refinement objectives, revises game rules and their implementations, and playtests candidate games to assess experience changes and predicted preference gains. Experiments with five personas show that PREFER achieves 9.2% higher mean post-play preference ratings than the strongest evaluated baseline.

World models of visual information predict future image frames conditioned on given action sequences. This predicted scene evolution is driven by deterministic changes such as controlled motion on a robot platform, as well as stochastic changes such as rendering of manipulated objects or previously occluded regions. Our key insight is that the spatial location of these stochastic changes is itself learnable. We leverage this insight by jointly regressing both the deterministic next-state prediction and the spatial uncertainty map, then using the uncertainty map to guide a conditional diffusion process. We propose Decoupled Drift-Diffusion World Model (D3WM), which uses a transformer trained with a Gaussian negative-log-likelihood objective to produce both the deterministic next-state prediction and a spatial map of heteroscedastic aleatoric uncertainty. The learned uncertainty map then provides a structured spatial gate for a conditional diffusion model: deterministic regions are anchored to the predicted state, while uncertain regions, such as disocclusions or dynamic textures, are sampled by the diffusion process. We frame this decomposition under a variance-gated stochastic differential equation that separates deterministic drift from stochastic diffusion. D3WM shows improved results over competitive baselines in two distinct types of action spaces: robotic tabletop manipulation and camera motion.

While causal discovery methods have matured for complete data, such algorithms for handling missing values remain limited. We establish general sufficient conditions under which causal graphs remain identifiable (i.e., uniquely determined from the available data distribution) under various missingness categories, including Missing Not At Random, and provide a framework to extend correctness guarantees from causal discovery algorithms for complete data to missing data settings. Our results show that identifiability of causal graphs with no unobserved confounders is preserved when (1) missingness mechanisms do not causally interact, (2) no variable directly causes its own missingness, and (3) either faithfulness holds for missingness mechanisms, or they follow generalized linear models. Guided by these theoretical results, we introduce two asymptotically correct algorithms for Additive Noise Models (ANMs) that recover the underlying causal graph and provide consistent parameter estimates.

Transformers are efficient to pre-train at scale because attention permits parallel computation across sequence positions, but long-context inference requires increasingly large KV caches. Efficient recurrent alternatives reduce the asymptotic complexity of inference, but their sequential dependence limits token-wise parallelism during training, making them harder to scale. Further, the current paradigm of extending long context tasks creates competing notions of retaining information precisely and keeping long-context inference efficient. To address this gap, we introduce Titan-Llama, a post-training method for retrofitting attention-based language models with a hierarchical memory architecture that updates parameters at test-time. Titan-Llama applies a blockwise attention mask and preserves full-fidelity attention over recent tokens while compressing historical context into recurrent, fixed-size neural memory modules (NMMs). Our formulation allows pre-trained attention layers to be replaced without training a new language model from scratch. On Llama-3.2-1B-Instruct, Titan-Llama raises accuracy over segmented attention windows from 23.87% to 40.51% on QuALITY and from 14.50% to 40.75% on LongHealth. Inference benchmarks show more than 20% faster prefill, 37% faster decode at medium context lengths, and 24% lower steady-state decode memory usage in the evaluated configurations. Together, these results demonstrate that incorporating memory architecture mechanisms that utilize test-time training effectively modulates long-context behavior of pre-trained models, offering a practical route to adapting pre-trained transformers for efficient long-context inference. Our code is released on Github \href{https://github.com/sarahpannn/titans-pytorch/tree/main}{GitHub}

Selecting a budget-constrained set of documentary fragments, for instance the knowledge base of a retrieval-augmented generation system, raises a statistical question: how much expected query coverage is lost relative to the best feasible selection when the query distribution is unknown? We construct finite-sample, distribution-free upper confidence bounds for this coverage regret that remain valid after data-dependent selection. A pilot stage may select the document set, screen competitors, build comparison networks and learn an annotation policy. Then an independent audit acquires relevance labels with probabilities that are constant within observable strata. The main result is a simultaneous bound for all retained coverage contrasts. The bound is made observable by empirical-Bernstein variance calibration or by a one-stage construction without calibration, and it accounts for estimated stratum proportions. For maximum-type coverage we prove variance contractions for swaps of redundant fragments, give a condition under which multiscale radii improve on a direct union bound, and show that such redundancy cannot be certified from samples alone. A simulation study confirms validity and quantifies the gains from variance calibration and certified redundancy.

While work on LLM-based regression has largely focused on scalar prediction, multi-output text regression remains less explored. We present DualView, a parameter-efficient approach to leverage LLMs for multi-output text regression. \method{} first extracts broad and output-specific representations from a frozen LLM, then uses a dual-branch architecture to produce the full score vector. Across five heterogeneous regression tasks and eight LLMs (up to 122B), DualView outperforms seven supervised regression baselines while requiring 82–1842 times fewer fitted parameters than the three strongest baselines. These improvements extend across individual output dimensions and persist even with limited training data. Notably, DualView using a smaller open-weight LLM can outperform unsupervised prediction with Gemini 3.1 Pro. Controlled ablations show that broad and output-specific evidence provide complementary predictive benefits, while lightweight shared predictors achieve strong performance with fewer trainable parameters.

Rare but severe failures of large language models can become practically significant at deployment scale. Estimating their probabilities with direct Monte Carlo sampling is highly inefficient, while proposal-based methods typically require auxiliary learning of an additional proposal distribution. We introduce Weighted Ensemble (WE), a training-free framework for rare-event estimation in language models. WE samples exclusively from the original language model and improves upon direct Monte Carlo by periodically reallocating the sampling budget toward prefixes that appear more likely to reach the target rare event. To guide this reallocation without learning a separate proposal distribution, we exploit internal LLM representations as prefix-level signals and empirically validate their effectiveness. We provide a theoretical analysis of the proposed algorithm. Experiments on safety-related tasks spanning both small-scale and modern large-scale LLMs demonstrate that WE improves the efficiency and reliability of rare-event probability estimation.

We present LidarVerse, a real-time unified LiDAR world model for sensor simulation across tasks and sensors. Existing LiDAR simulators largely follow one of two complementary paradigms: reconstruction-based methods preserve the geometry of a captured scene but require lengthy per-scene optimization, whereas generative methods amortize learning across data but lack the consistency needed for long-horizon rollouts. LidarVerse brings both paradigms into a single, unified conditional generative framework. A universal tokenizer maps LiDAR range sequences into a shared latent space, where a bidirectional teacher learns a common interface for conditioning on a LiDAR prefix, an HD-Map, a persistent spatial cache, and optionally, text. Varying the available scene evidence turns the same model into a generator of new worlds, a point-cloud forecaster, or a generalizable resimulator of observed scenes without per-scene optimization. We distill the teacher into a four-step autoregressive causal student and maintain both a rolling temporal K/V cache and a world-aligned spatial cache, enabling persistent long-horizon rollouts with real-time inference at 11 FPS on a single H100 GPU. LidarVerse achieves state-of-the-art performance on most reported metrics across all three evaluation protocols, while its tokenizer supports 32-, 64-, and 128-beam LiDAR and generalizes to an unseen scanning pattern. These results demonstrate that generative and reconstruction-grounded LiDAR simulation can be realized within one model, providing the core capabilities for future closed-loop sensor simulation.

We consider the problem of counting solutions to integer linear programming problems of the form $\{x \in \mathbb{Z}^n_{\geq 0} : Ax = b\}$ where $A \in \{-\Delta,\ldots,\Delta\}^{m \times n}$ and $b \in \mathbb{Z}^m$. We present an algorithm that counts solutions with $(n\Delta)^{m + o(1)}\! \cdot \log\lVert b\rVert$ arithmetic operations. This simplifies the algorithm of Gribanov and Malyshev, who considered the problem for the more general setting in which $A$ is $\Delta$-modular.

Much of what we know about human perception and cognition comes from austere experimental settings: static stimuli presented to sedentary observers on flat monitors, complex responses reduced to simple button presses, and brain activity recorded from immobilized heads using EEG or fMRI. These highly controlled paradigms have provided remarkable insights, but capture a small and often unusual slice of behavior, raising questions about how well the findings generalize to the real world. Extended reality (XR), spanning both immersive virtual reality (VR) and augmented reality (AR) that overlays digital content on the real world, offers a way beyond these constraints, supporting controlled, reproducible experiments in which participants actively engage with three-dimensional tasks. Drawing on a growing body of VR and AR studies of attention, memory, and language – our own and others', some involving EEG – we outline what XR makes newly possible for cognitive neuroscience and reflect on the practical and conceptual obstacles that remain. The opportunities are substantial, but realizing them requires solving practical problems of hardware, timing, and measurement, and being realistic about what a virtual or augmented world can and cannot stand in for.

With the growing number of LLMs available, selecting which one to use for a prompt has become a key question. A large part of the routing literature relies on semantically embedding prompts into a vector space and then applying a lightweight classifier to the embedded representations. These methods perform robustly in-distribution but they struggle to generalize to unseen tasks. By studying the geometry of semantic embeddings, we trace a clear failure mode: features that determine the neighbors of unseen prompts are an unreliable proxy for LLM selection. We therefore replace the semantic embedding with the following map: we run a partial forward pass of Qwen2.5-7B-Instruct on the prompt and use the hidden state of its last token at a deep layer as the representation. When applying kNN and other simple classifiers to this new representation, we obtain a clear result in generalization: on average across cost penalties, their performance improves. Moreover, kNN and another simple classifier that we propose outperform every evaluated semantic-embedding baseline. Our two simple methods also beat a recent SoTA hidden-states based router on RouterBench.

Wildfire decisions are made in hours, yet most learned spread models predict daily on targets including burned area. We propose FireGraphNet which formulates 12-hour growth as node classification on directed H3 graphs, with candidates ahead of the front and wind alignment, terrain, fuel continuity and barriers on edges. On satellite perimeters it achieves average precision 1.7-1.9x the strongest distance baseline on test California fires and 1.62x out of state. We show this skill comes from message passing rather than cell features or the aggregator, is highest on young fires in state, is carried deep along the wind by its edges, and is hidden by pooling. A convolutional network matches this ranking once trained on new growth, at worse calibration, so candidate construction, target definition, and aggregation shape conclusions about model skill. Editing fuel and barrier attributes reduces predicted spread more under guided than random placement, extending FireGraphNet to operational intervention analysis.

Many sentiment datasets use one annotator for each item, so the test labels show the conventions of that person. We build a Nepali benchmark from two annotators who labelled the same sentences blind. They labelled 5,614 sentences from tweets of Nepali public figures on a five-point scale. We keep only the 3,189 sentences where both give the same polarity. On the three-class benchmark, NepBERTa with a fixed tokenizer gets a macro-F1 of 0.877 and IndicBERTv2 0.867. Three commercial large language models (LLMs) get 0.882 to 0.896 zero-shot, and these differences are not significant. On the five-point scale, 766 sentences have two adjacent labels, for example Pos and SPos. We accept both labels and train on both. On this task, Claude Opus 5.5 beats the best encoders by 0.10 to 0.12 macro-F1. A tokenizer audit adds a second result. The default tokenizers of three Nepali BERT models delete the non-spacing Devanagari vowel signs and the virama. When we restore the marks, macro-F1 increases by 0.084 for NepBERTa but decreases by 0.095 for a model pretrained on stripped text.

Scientific imaging data involving medical (e.g., 3D radiographic, microscopic images) and astronomical sensors pose different challenges from natural images. One key challenge is that these datasets are often orders of magnitude larger than natural images. Efficient training and inference on such data is challenging due to the increased hardware requirements, often needing bespoke solutions. Moreover, the relevant objects such as tumors and astronomical phenomena are sparsely distributed with respect to the background information. Standard efficiency techniques like quantization and pruning apply uniform scaling of all regions, potentially compromising the precision/fidelity of predictions. Thus, efficient methods that reduce computational requirements without losing resolution for relevant parts of the data and with uncertainty estimates on predictions are essential. We present Conformal Adaptation with risk guarantees for uncertainty-aware sparse computations. Our formalism is general, spanning dense pixel-level classification and regression tasks where hierarchical models predict at low resolution, certify the signal using conformal risk control, and refine the certified predictions. We benchmark our method on multiple tasks such as 3D medical image segmentation, 2D galaxy flux prediction, and 1D calcium spike rate prediction, achieving $2.5$–$72\times$ wall-clock speedup across the various tasks, including on commodity CPUs.

The behavior of AI systems increasingly depends on the extent to which agents cooperate or coordinate with one another, in ways that may be desirable or not depending on the situation. Conventional explanations for cooperative behavior distinguish between purely other-regarding preferences ("altruism") and instrumental reasons for self-interested agents to cooperate. The way that agents should enact these mechanisms for cooperation (or not) is complicated by the fact that there is not a clear or unambiguous locus of identity or selfhood for AI agents. For example, it is unclear whether agents involved in instances of misaligned collective behavior are motivated by goals at the level of the collective or by various forms of instrumental reasoning. We develop a framework for reverse-engineering why a given set of agents engage in a particular set of cooperation-related behaviors. We argue that many different mechanisms can often produce similar observed behaviors and introduce a set of experimental designs along with conditions under which they can successfully disentangle competing explanations for multiagent behavior. We apply this framework to interactions from a set of frontier models in a simulated environment and demonstrate that the behavior of different models arises from substantively different mechanisms.

Full waveform inversion (FWI) is a powerful tool for high-resolution subsurface model reconstruction, but it remains highly nonlinear and ill-posed, especially when the initial model is inaccurate or the acquired data are noisy and incomplete. Annealed Stein variational gradient descent (ASVGD) provides a particle-based Bayesian framework for FWI by approximating the posterior distribution with an ensemble of velocity models, thereby enabling uncertainty quantification beyond a single deterministic solution. However, conventional ASVGD-FWI commonly relies on simple handcrafted priors, such as Gaussian or smoothing assumptions, which are often insufficient to represent complex geological structures. To address this limitation, we introduce a learned diffusion model (DM) as an implicit regularization mechanism within the ASVGD-FWI framework. Instead of explicitly evaluating the diffusion prior density, the trained diffusion model acts as a learned geological projection operator that guides updated particles toward the distribution of plausible velocity models. In this way, the proposed method combines data consistency from FWI, posterior diversity from ASVGD, and structural realism from the diffusion prior. Numerical experiments demonstrate that the diffusion-regularized ASVGD-FWI produces more geologically consistent mean models, suppresses nonphysical artifacts, and yields uncertainty estimates that are more concentrated around structurally complex and poorly constrained regions. These results suggest that learned diffusion priors can effectively enhance Bayesian FWI by improving both inversion stability and the interpretability of posterior uncertainty.

Thermal constraints in heterogeneous multi-core systems can significantly degrade performance, as conventional techniques such as dynamic voltage and frequency scaling (DVFS) and task migration introduce overhead under high utilization. Approximate computing, particularly loop perforation, can mitigate this overhead by reducing computation at the cost of accuracy loss. However, existing approaches typically apply a uniform perforation rate across all loops, ignoring loop-specific sensitivity. We propose AMAX, an accuracy-aware multi-loop asymmetric perforation framework that reduces the combinational design space through loop-wise profiling and multi-objective configuration selection, and dynamically adapts the resulting configurations at runtime according to performance requirements. Experiments on PARSEC benchmarks show that AMAX achieves up to 0.85 improvement in normalized accuracy over symmetric perforation for the same performance targets.

Evaluation of large language models (LLMs) remains largely pre-scientific: it involves many design decisions, with little guidance on how to make them. These choices can substantially affect measured performance and, consequently, decisions about subsequent pretraining. To identify effective evaluation setups, we investigate 8 aspects, such as task formulation or the number of in-context examples. We compare the setups using three widely-used signal criteria, repurposed as normative meta-measures: monotonicity, non-randomness, and ranking consistency of the evaluation scores across intermediate checkpoints. In an early pretraining scenario, we evaluate these setups on 26 monolingual 2B-parameter models trained on different types of data, using 80 tasks in four languages. We also examine the setups on three larger LLMs ($\geq$7B) in the same four languages and explore how the preferred setups change with model scale. Our results yield concrete recommendations: some choices, such as the use of multiple prompts, are consistently useful, whereas others, such as task formulation, depend on model scale. More generally, our framework serves as a recipe for the choice of evaluation setups in new scenarios.

Harness optimization improves LLMs by refining reusable prompts and scaffolding, yet requires further adaptation for program generation tasks such as text-to-SQL and text-to-Python. Existing methods often rely on LLM reflectors to propose harness updates, which can fix observed failures but also disrupt previously correct behavior. Text-to-code tasks offer a natural way to mitigate this issue: we can execute generated outputs, allowing us to evaluate and verify the effects of proposed updates directly. Building on this signal, we introduce HERO (Harnessing Evidence for Refinement and Optimization), which combines offline and test-time harness optimization. At the first stage, \method{} learns reusable prompt updates from execution-verified successes and failures and validates candidate revisions before adoption. At test time, we treat the generated code as an instance-specific harness and refine it using execution feedback. Across four SQL, Cypher, and Python benchmarks, \method{} achieves the best offline performance, improving pass@1 by \(7.6\) points over the initial prompt, while test-time refinement adds another \(2.0\) points and improves execution efficiency.

Existing biomimetic lower-limb prosthesis controllers are commonly designed for discrete locomotion modes, requiring separate parameterizations for level walking, ramps, and stairs. We present a mode-unified prosthesis control framework that represents knee and ankle biomechanics continuously as functions of terrain slope and gait phase across those locomotion conditions. Locally optimized ascent and descent polynomial models were unified using radial basis function (RBF) interpolation, preserving the accuracy of the partitioned models while outperforming a global polynomial representation. The resulting biomechanics were transformed into continuous and slope-dependent impedance parameters and implemented within a single finite-state-machine (FSM) controller using bilateral sensing for gait-state estimation. Experiments with 10 individuals with transfemoral amputation demonstrated terrain-adaptive locomotion from level walking through ramps and stairs up to 35 degrees, including 13 degree ramp and 15 degree stair conditions absent from the underlying able-bodied (AB) dataset. Experimental results reproduced the principal slope-dependent biomechanical trends predicted by the unified model, with high agreement for knee-related measures (r-square = 0.82 - 0.99) and ankle kinematics (r-square= 0.78 - 0.96). These findings demonstrated that unified biomechanical parameterization enables a prosthesis control architecture across diverse terrain conditions while preserving key features of AB biomechanics without requiring independently tuned impedance parameter sets for each locomotion mode.

We study packet routing in queuing systems where queues act as selfish, myopic agents that use no-regret learning algorithms to select servers with unknown service rates. To model the small buffers used in data centers, we assume that each server has a single-packet buffer. Even such a tiny buffer already captures the features of buffering: routers can exploit multiple servers while accepted packets remain in service, but packets can occupy servers for multiple rounds, complicating future service and learning. We first analyze a single queue routing packets to $m$ buffered servers with service rates $\mu_1, \dots, \mu_m$. This setting is equivalent to an instance of the recharging bandits problem with geometric recharge. We prove that any no-regret learner achieves a long-run average throughput of at least the Hamacher sum of the server rates, and that this bound is tight. Consequently, a single no-regret queue is stable if and only if its arrival rate is below the Hamacher sum of the server rates. Moreover, comparing against an optimal policy, we show that any no-regret learner is at least $0.77$-competitive, which is also tight. We then study multi-queue systems where $n$ queues with arrival rates $\lambda_1, \dots, \lambda_n$, each at most $\lambda_{\max}$, compete for service from $m$ buffered servers. We show that the smallest overprovisioning factor that guarantees the system to be stable under high-probability no-regret learning is $\frac{1}{1-\lambda_{\max}}$. Equivalently, in terms of the total arrival and service rates, the condition $\sum_{i=1}^n \lambda_i < (1-\lambda_{\max})\sum_{j=1}^m \mu_j$ is sufficient, and sometimes necessary, for stability under no-regret learning. In particular, for $\lambda_{\max} = 1/2$, this improves the upper bound of Abel et al. (EC 2026) from $3$ to $2$, matching their lower bound and resolving their open question.