Papers

Large language models (LLMs) typically generate text autoregressively (AR), predicting one token at a time. Block diffusion language models (dLLMs) instead generate blocks sequentially while denoising multiple tokens in parallel within each block, offering a promising way to accelerate generation. Rather than training such models from scratch, recent work adapts strong pretrained AR models into block dLLMs through distillation. On-policy distillation (OPD) has been widely used for LLM training because it supervises the student on states generated by its current policy, rather than only on fixed offline trajectories. By training on the states the student actually visits, it reduces the mismatch between training and generation and can provide more relevant supervision as the student evolves. Recent work has extended this idea to AR-to-block-diffusion conversion. However, this setting introduces a fundamental mismatch in supervision: the block-diffusion student and the causal AR teacher condition on different information at the same training state. The student predicts from the entire partially denoised block, including visible future context, whereas the standard AR teacher target is defined only from the causal prefix. As a result, the teacher distribution used for distillation is not fully aligned with the information available to the student. We therefore introduce d-OPD, a future-aware on-policy distillation method that corrects the AR teacher distribution to better align with the student-visible state by incorporating visible future information within each block, providing supervision that better matches the information used by the student. Across Qwen3 models from 0.6B to 8B, d-OPD improves the six-benchmark average by up to $4.0$ points over OPDLM and reduces training time by $1.35$-$1.58\times$. The code is available at https://github.com/mit-han-lab/d-OPD.

The ongoing transition toward quantum-safe networking has motivated the development of hybrid network architectures integrating Quantum Key Distribution (QKD) and Post-Quantum Cryptography (PQC). However, the experimental evaluation of hybrid QKD-PQC network architectures remains constrained by the high cost and limited accessibility of quantum hardware, as well as by the limited support for hybrid QKD-PQC networks in existing emulation platforms. Quditto is an open-source emulation platform originally designed for QKD networks that enables cost-effective and reproducible experimentation without requiring dedicated physical quantum infrastructure. Building on this foundation, this work presents Quditto as a hybrid QKD-PQC network emulation platform featuring automated and scalable cloud-native orchestration. The proposed platform introduces four principal contributions: a cloud-native orchestrator enabling fully automated infrastructure deployment across cloud and multi-cluster environments; an optimized provisioning workflow enabling large-scale quantum-safe network emulation; native integration of post-quantum nodes enabling unified emulation of hybrid QKD-PQC networks; and a secure key management module providing persistent and access-controlled storage of cryptographic material. Experimental validation demonstrates sublinear orchestration-time scaling with network size and successful end-to-end hybrid QKD-PQC key establishment on a representative spine-leaf deployment, thereby enabling the systematic evaluation of quantum-safe networking mechanisms in large-scale heterogeneous network environments.

Coding agents are increasingly deployed for iterative development on real repositories, yet existing evaluation barely answers a basic question: do coding agents reuse existing code or reinvent the wheel? The question matters: every duplicated implementation is a fix applied twice and agents produce code far faster than humans can audit, so redundancy accumulates unsupervised. Thus, we present RepoReuse, a multi-turn benchmark for auditing code reuse in real repositories, where requirements are revealed turn by turn and the workspace accumulates across turns. It is built by a fully automated pipeline combining AST-based dependency graphs, guided evidence collection, and execution-verified task synthesis, and scales readily to new repositories. Beyond pass rates, we measure the reuse rate together with recall and cross-turn structural redundancy. An audit over 3{,}000 turns shows that agents progressively stop exploring relevant repository code, reuse their own history less even when it is fully in the workspace, and leave duplicated logic in 50.8% of task chains by turn~5—all while pass rates barely move. Such deficiencies are invisible to pass rates, underscoring the need to evaluate code generation beyond functional correctness.

Supervised fine-tuning can teach language models undesired behaviours alongside desired ones. Inoculation prompting (IP) aims to limit unwanted generalisation by requesting the undesired behaviour during training and removing the request at inference. However, undesired behaviour can still appear under unrelated prompts. IP can also hinder learning of the desired behaviour. We address these limitations in settings where both behaviours co-occur in most training examples, so filtering out examples with undesired behaviour leaves only a small clean subset. We introduce stratified inoculation prompting (SIP). SIP leverages a small clean subset to demonstrate that desired behaviour should persist without the undesired one across different contexts. SIP oversamples these clean examples under diverse non-eliciting prompts while inoculating the rest. SIP substantially reduces expression of undesired behaviour while preserving more of the desired behaviour than IP. These gains persist even when we extend IP to oversample the same clean subset at the same rate as SIP. Moreover, SIP yields lower emergent misalignment rates in all harmful-advice setups we tested. SIP can be further extended to limit the undesired behaviour even under prompts that explicitly request it. We introduce backdoor dilution, which weakens expression under the inoculation prompt, and password-locked inoculation, which concentrates elicitation on a designated password. Taken together, our findings show that changing the training contexts for a small clean subset can significantly improve selective generalisation.

Fully homomorphic encryption (FHE) enables neural network inference directly on encrypted inputs, but it remains orders of magnitude slower than plaintext in- ference. Applying the server's plaintext weights to encrypted activations involves plaintext-ciphertext multiplications (PMult) and accounts for more than half of inference time in recent systems. Ternary quantization can replace these multipli- cations with additions and subtractions, but the savings rarely materialize under packed execution. A single PMult applies a weight group fixed by the packing layout and can be avoided only when all its weights share the same ternary value. Ternarizing all groups, however, largely degrades accuracy. We present FIONA, an offline optimizer that selectively ternarizes weights within a given packing layout based on the estimated effect of ternary conversion on the model's performance. FIONA encourages a shared ternary value within each weight group and retains full-precision weights for sensitive groups, so ternar- ized and full-precision paths coexist within a layer. It then compiles these hybrid operators exactly, applying common scaling factors once to accumulated inputs and reusing sums across outputs. Weight ternarization can also narrow the input ranges of downstream polynomials. FIONA fits lower-degree replacements under a cumulative accuracy budget, reducing multiplicative depth and bootstrapping. On VGG11, ViT, and BERT, FIONA reduces PMult operations by 53.4-79.5% and accelerates end-to-end encrypted inference by 2.38x, 1.68x, and 1.84x, re- spectively, with less than 1% accuracy loss across all three models.

Interference is commonly treated as geometric overlap between learned features. We introduce effective interference, which combines feature geometry and code statistics to capture realized interactions, distinguishing constructive from destructive interference and frequent weak interactions from rare strong ones. Under local fixed-support assumptions, we characterize how architectural constraints shape interference through four mechanisms: feature orthogonalization, bias compensation, gain adaptation, and encoder-decoder separation. Experiments with sparse autoencoders show that constrained architectures selectively reduce overlap among co-active features, while bias, gain, and encoder freedom allow constructive cross-contributions to remain. Together, these results show that interference in learned representations depends not only on feature geometry, but also on how features are used and on the architecture that produces their codes.

While Large Reasoning Models (LRMs) excel at complex reasoning, alignment through reinforcement learning often induces systemic overconfidence. In production environments, where logits may be unavailable, robust black-box uncertainty quantification (UQ) is essential for trustworthiness and safety. Focusing on question-answering for LRMs, we show that existing black-box methods, such as paraphrase-based self-consistency and confidence verbalization, offer little to no improvement over simple repeated sampling, suggesting that alignment suppresses useful output variability. We introduce prompt-level relaxation operators that broaden the model's effective output distribution by approximating the effect of an optimal policy obtained with a stronger KL-regularization parameter, hence closer to the reference model. Theoretically, we demonstrate that relaxation improves calibration. We propose Jailbreak for Uncertainty (J4U), a jailbreak-derived technique for UQ that empirically reproduces the behavioral signatures predicted by our relaxation theory. Across 3 datasets and 4 LRMs, including a closed-source production model, J4U's improvement over repeated sampling achieves statistical significance in up to 6 times more LRM-dataset-metric settings than the strongest black-box UQ state-of-the-art baseline we evaluate, with average ECE reductions up to 5 times larger. These results provide a practical tool for UQ in black-box LRM deployment.

The evaluation cost of transformers with softmax attention scales quadratically with sequence length. Kernel attention addresses this by replacing softmax with a more general kernel function. In this paper, we aim to identify kernels that retain the expressivity of attention while enabling quasi linear computation. To quantify expressivity, we introduce a capacity for each kernel, measuring the maximum sequence length for which the attention matrix can approximate the identity. A higher capacity thus indicates greater expressivity. We show that expressive kernels like softmax, Gauss, and Laplace have infinite capacity. In contrast, common quasi linear kernels, such as those derived from finite dimensional feature maps, exhibit finite capacity. As a solution, we propose additive kernels constructed from univariate spline and polynomial exponential kernels. We prove that these maintain infinite capacity while allowing quasi linear computation via sorting. Finally, we implement additive sorting kernels efficiently and benchmark them against modern softmax backends, demonstrating advantages for long sequences.

Estimating probability densities from a finite set of samples typically requires dataset-specific model fitting. We introduce PRODiGI, a pretrained data-to-program model that infers an explicit, executable generative program in a single forward pass. Pretrained on synthetic datasets paired with their ground-truth programs, PRODiGI accommodates diverse generative families and data dimensionalities through template prediction and non-autoregressive program parameter decoding. Its inferred programs support direct sampling, density and score evaluation, and inspection independently of the pretrained model. We further introduce program-space fine-tuning, which refines differentiable program parameters by matching generated and empirical samples while keeping model parameters intact. Experiments show that PRODiGI achieves lower average density and score MAE than existing pretrained models, while offering multi-fold speedups over its closest competitors. Program-space fine-tuning further reduces generation MMD by 84%. By turning empirical data into explicit, reusable programs, PRODiGI introduces a new direction for fast, interpretable tabular generative modeling.

Reinforcement learning can turn one language model into several specialists, each excellent at a single skill such as mathematics, coding or following instructions, but users need one model with all of these skills. Multi-teacher on-policy distillation (MOPD) merges them by letting the specialists teach one student: the student answers each prompt, and the specialist for that prompt's domain gives feedback on every token. This routing decides which specialist teaches, but not how strongly its feedback moves the shared student. In Qwen3.5 models at three sizes, we find that MOPD's student does not beat one taught by the best single specialist and gains little of the mathematics specialist's advantage. The feedback is unbalanced: instruction-following feedback is several times more spread out than mathematics feedback and dominates the student's updates. We propose Domain-Normalized MOPD (DN-MOPD), which keeps the routing and rescales each domain's feedback by its measured spread. On six public benchmarks, DN-MOPD improves the average score over MOPD at every size, across three random seeds and under two answer-length limits, and recovers most of the lost mathematics gain. Controls with fixed domain weights show that the gain comes mainly from turning down instruction-following feedback rather than turning up mathematics alone, and that fixed weights close to those DN-MOPD measures perform comparably. Combining specialists therefore requires deciding not only which one teaches, but also how strongly its feedback counts.

Large audio-language models (LALMs) excel at understanding and reasoning tasks over atomic sound events, yet their ability to infer higher-level human activities from such fine-grained events remains largely unexamined. Everyday human actions and activities, such as setting a table, cleaning the house, or preparing a breakfast emerge compositionally from temporally distributed sound events, requiring abstraction beyond the event-centric granularity that dominates current training and evaluation paradigms. Our benchmark evaluates a wide set of LALMs under a principled framework that tests how language-based reasoning, grounded in acoustic perception, structures sound abstractions into higher-level understanding. By systematically varying exemplar typicality and distractor similarity, our evaluation exposes \added{that current models do not reliably perform compositional inference from atomic acoustic events to higher-level human activities solely from audio.} All data, taxonomies, and evaluation scripts are publicly available on our companion website: https://alm-sounding-actions.onrender.com/

The appearance of Jev marked the era of System One Models, foundation models that return structured decisions with probability distributions rather than text. Aside from low cost and great speed, Jev's central promise is that these probabilities are calibrated: such claim is not backed by any public test and available external benchmarks evaluate confidence calibration, not whether every returned option probability has the right numerical meaning. To tackle this issue, we introduce Sys1Cal-v1, a dataset of True/False questions about a proposition $A$ for which the exact probability $P(A)$ is known by construction. Each item is queried through the three Jev primitives - Noul, Choice and Score - and evaluated by total variation distance from the ground-truth distribution, which can be used to estimate a soft accuracy of System One Models. We showcase the utility of Sys1Cal-v1 as a benchmark dataset by evaluating Jev and SemIf, an open-source Choice-style baseline. In this work, however, we focus even more deeply on Jev, by studying the calibration of its Score and Choice answers. In particular, we discover a peculiar behaviour that can be explained by assuming that Jev suppresses a third truth value, going beyond True and False. In other words, in Choice answers, $P(A)$ and $P(\neg A)$ are presented as if $P(A)+P(\neg A)=1$, while a term $P(U)\neq0$ is missing in the sum. Recovering $P(U)$ leads to an improvement of median soft accuracy in Choice answers from $0.771$ to $0.978$, suggesting that, even in binary decisions, Jev wants to answer with a third option:“I don't know”.

Evaluating semantic similarity between videos is a fundamental challenge in computer vision, essential for tasks ranging from out-of-distribution (OOD) detection to video retrieval. However, defining and labeling video similarity is notoriously difficult and expensive due to the complex spatio-temporal nature. In this paper, we propose a novel self-supervised approach that leverages generative uncertainty from text-to-video (T2V) diffusion models to learn semantic similarity without human annotations. Our method is based on the observation that T2V models produce consistent outputs for familiar concepts but exhibit high variance and uncertainty when prompted with specialized concepts. We utilize this behavior to identify stable semantic features within existing pretrained representations, such as VideoMAE and V-JEPA. Specifically, we learn a mask over these embeddings using purely generated data, encouraging the model to retain features that remain consistent across generations of general concepts while discarding those associated with generative noise or uncertainty. Experimental results across three key tasks demonstrate that our learned feature subspaces consistently outperform original pretrained features and baseline feature selection methods.

Zero-knowledge proofs (ZKPs) enable the verification of a statement without revealing any information beyond its validity and constitute a fundamental primitive in cryptography and information theory. However, existing constructions rely on computational assumptions and are predominantly confined to bipartite settings, leaving their information-theoretic realization in bipartite or network scenarios largely unexplored. Here we develop a framework for zero-knowledge verification based on the indistinguishability of quantum states under operational constraints. Exploiting the fundamental limitations imposed by local operations, we show that a verifier is inherently restricted from extracting information about the underlying state while retaining the ability to verify correctness. We construct explicit protocols for multiparty quantum networks that achieve information-theoretic security, ensuring that no subset of collaborating parties can gain knowledge beyond the validity of the statement, independent of their joint computational power.

Mechanistic discovery in neuronal microenvironments requires interventions and measurements that separate competing explanations of solute transport and neuronal response. Predictive accuracy cannot settle the question: a real mechanistic change and an error in the computational twin leave the same signature in sparse observations. We formalize this twin confounding and reason over a joint mechanism–discrepancy belief, designing experiments that separate the two. NeuronDiscover is an Agent-in-Twin framework whose shared, mechanism-grounded World Action Model (WAM) couples prediction, intervention proposals, and observation design; independently adjudicated outcomes revise a scoped Mechanism–Intervention–Observation–Outcome (MIOY) graph, whose supported relations compile into executable programs carrying discrepancy-adjusted acceptance bounds. We evaluate on simulated brain-fluid tracer-transport worlds adjudicated by an independently frozen finer-mesh reference solver, and on donor-disjoint public current-clamp recordings of cortical neurons. Counting only relations that reach a certified terminal status, and scoring abstentions as unresolved for every method, at a matched budget of 16 experiments over 32 source units NeuronDiscover resolves 4.0 relations per assigned world against 3.4 for the strongest baseline and 3.2 without graph revision, at 5% false support and 82% scope accuracy. Joint mechanism–discrepancy acquisition resolves 3.8 relations versus 2.9 for plug-in expected information gain; discrepancy-adjusted verification lowers accepted-program failure from 15% to 9% at 60% acceptance coverage; and transfer to the recordings yields 1.94 versus 1.53 relations per assigned world. Correctness is adjudicated within declared model worlds and archival recordings.

Controlling tail risk is an important objective in online optimization, and recently it has been studied in the context of competitive analysis. Continuing this line of research, we investigate pure tail constraints, which capture the tradeoff between expected and worst-case competitiveness. For two fundamental search problems, online bidding and line search, we derive the Pareto-optimal frontiers of this tradeoff. We then investigate another classic problem, TCP acknowledgment, which has structure similar to the iterated ski rental problem. There, we construct an algorithm whose tradeoff coincides with the known Pareto-optimal tradeoff for ski rental. The lower bounds for this problem are substantially more involved as the problem exhibits adaptive structure: an online algorithm observes requests of the adversary (packet arrivals) and may adaptively adjust its actions (acknowledgments) on this basis. We emphasize that all previous work on tail risk in the context of competitive analysis was restricted to non-adaptive problems, where the feedback given to an algorithm was essentially limited to a binary indicator of whether the algorithm has succeeded or not. Nonetheless, we identify a set of constraints implied by tail bounds in this adaptive setting, and show that they imply a nontrivial lower bound on the TCP acknowledgment problem.

Despite the promise of Large Language Models (LLMs) in computational chemistry, rigorous combinatorial chemistry problems remain difficult because they require quantitatively constrained molecular modification, candidate validation, and systematic revision after failed attempts. Existing tool-augmented chemical agents demonstrate useful planning and tool use, but they rarely provide a unified loop for property-driven molecular optimization and workflow-level composition. To bridge this gap, we propose Tool-Grounded Multi-Agent Reasoning for Compositional Chemical Problem Solving (TMCS), a step-by-step multi-agent framework that formalizes chemical problem solving as an interpretable, tool-augmented workflow. At the task level, specialized agents leverage external tools, few-shot trajectory memory, and structured reflection to iteratively refine solutions. At the workflow level, TMCS chains generation, understanding, editing, description, and optimization into a closed-loop pipeline. Evaluations across multiple chemical tasks demonstrate that TMCS consistently enhances chemical reasoning across both open- and closed-source base models, achieving state-of-the-art performance.

Machine learning task type identification is essential for constructing valid ML pipelines, yet in practice it is typically specified manually. We investigate whether large language models (LLMs) can infer both the data domain and the downstream prediction task directly from dataset-level information when only the target feature is provided by the user. Together with our LLM-based system we also release an annotated benchmark comprising 625 public tabular and time series datasets. We evaluate the proposed approach in three settings: (i) tabular datasets in comparison with established AutoML heuristics, (ii) cross-domain evaluation across tabular and time series datasets, and (iii) a practical deployment scenario using smaller local models. The results show consistent advantages for LLM-based task type identification, with increasing difficulty in heterogeneous and resource-constrained settings. LLM-based approaches outperform AutoGluon in the tabular setting, reaching 0.98 F1 macro compared to 0.93. In the cross-domain setting, the best model achieves 0.90 F1 macro, while smaller locally deployable models reach 0.75, indicating a trade-off between deployment feasibility and accuracy.

Algorithm Distillation (AD) has demonstrated the remarkable ability of Transformers to perform in-context reinforcement learning without explicit weight updates. However, capturing long-term learning progress necessitates expansive context windows, which incur prohibitive memory costs and limit scalability in complex, long-horizon tasks. To address this bottleneck, we propose Recurrent Algorithm Distillation (RAD). RAD employs a dual-component architecture: a Compression Transformer that distills extended interaction histories into compact latent tokens, and an AD Transformer that auto-regressively generates actions using a hybrid context of these compressed memories and recent transitions. By maintaining a fixed-size latent buffer, RAD decouples the effective history length from computational complexity, functionally providing the model with a long-horizon memory. Empirical evaluations across diverse environments demonstrate that RAD matches the asymptotic performance of standard AD with significantly reduced context window sizes, offering a scalable solution for efficient in-context decision-making.

We study the Reverse Sequential Proportional Approval Voting Rule (RevSeqPAV) in approval-based committee elections. Despite its historical prominence and practical use, its properties and guarantees are much less understood than those of Sequential PAV. We analyze it along two dimensions: proportional representation (measured by Extended Justified Representation, its approximations, and proportionality degree) and approximation of the maximum PAV score of instances. We first establish strong negative results for general, unrestricted election instances and then identify settings in which the rule provides meaningful fairness and optimization guarantees.

Meta-Black-Box Optimization (MetaBBO) is one of the highlights in the recent AI for Optimization trend. This paradigm's bi-level workflow leverages the learnable algorithm design policy at meta level to ensure the performance and generalization improvement on the low-level optimization task. While MetaBBO helps advance the performance lower bound of the resulted optimization system, it is currently handcrafted and customized case by case to adapt different optimization problems, which inevitably introduces inherent subjectivity and hence restricts the performance upper bound and usability in practice. In this paper, we address this issue by regarding MetaBBO's design loop as coding task, where we could introduce openendedness into MetaBBO with recursive self-improvement capability of advanced coding agents. Specifically, we propose a dual-agent framework: i) a task agent continuously refines the codebase of a target MetaBBO approach through code evolution; ii) a hyper agent progressively modifies the task agent and itself to provide open-ended design behavior; iii) the evolved MetaBBO codebase is evaluated and all in-execution information is fed back to the agents for recursive self-referential improvement. As a result, given a naive MetaBBO template, our framework automates a design evolution and finds novel variants superior to up-to-date human-made MetaBBO baselines. Surprisingly, the experimental results also demonstrate that our framework supports fast adaption across different optimization domains. Solid interpretation analysis further reveals interesting design principles emerge in such open-ended process. This work serves as the first exploration on automating design of complex learning-assisted optimization algorithms.

A longstanding conjecture in quantum complexity theory asserts that, under the uniform input distribution, quantum query algorithms can be polynomially simulated by classical query algorithms. More precisely, the acceptance probability of any quantum query algorithm can be approximated, on average over uniformly random inputs, by a classical query algorithm, with only polynomial query overhead. The conjecture is central to understanding whether exponential quantum advantages for decision problems necessarily rely on additional structure. We study analogues of this conjecture under other natural input distributions and prove that they are all equivalent to the original uniform-distribution conjecture. We first consider the product distribution $\mu_p$, where the input bits are independent Bernoulli variables with fixed bias $p$. We show that for every fixed $p \in (0, 1)$, quantum query algorithms under the $\mu_p$ distribution admit polynomial-overhead classical simulations if and only if the same holds under the uniform distribution. Second, we consider the distribution $\nu_p$ that is uniform over the slice of strings with Hamming weight $\lfloor pn \rfloor$ and prove a similar equivalence for the $\nu_p$ distribution and the uniform distribution. The Aaronson-Ambainis conjecture is a stronger statement that implies the above-mentioned conjecture and is formulated in terms of bounded low-degree polynomials on the Boolean hypercube. It asserts that under the uniform distribution, any such polynomial with nonnegligible variance must have an influential variable. We formulate analogues of this conjecture, where the underlying distribution is a biased product distribution or a uniform distribution over a slice, and prove that all these variants are equivalent to the original Aaronson-Ambainis conjecture.

Using the recent exponential correlation bounds of Chattopadhyay, Hatami, Lee, Lovett, Tal and Viola between $\mathbb{F}_2$-polynomials and the XOR of majorities, we show that $\mathrm{Almost}\text{-}\oplus\mathrm{P} = \mathrm{BP}\cdot\oplus\mathrm{P}$, where $\mathrm{Almost}\text{-}\oplus\mathrm{P}$ is the class of languages that lie in $\oplus\mathrm{P}^R$ with probability one for a random oracle $R$. This is the parity analogue of Bennett and Gill's $\mathrm{Almost}\text{-}\mathrm{P} = \mathrm{BPP}$ and Nisan and Wigderson's $\mathrm{Almost}\text{-}\mathrm{PH} = \mathrm{PH}$. The key ingredient is a pseudorandom generator with polynomial seed length that fools $\mathbb{F}_2$-polynomials of polynomial degree on exponentially many variables. As an application we complete a random-oracle proof of the first half of Toda's theorem, $\mathrm{PH} \subseteq \mathrm{BP}\cdot\oplus\mathrm{P}$, following an approach of Regan and Royer. Relative to a random oracle, the polynomial hierarchy collapses into $\oplus\mathrm{P}$ by applying Valiant-Vazirani and Papadimitriou-Zachos level by level, with no probabilistic quantifier ever moved through an oracle. Our result then removes the oracle. We compare this argument with the simple proof of Toda's theorem by Fortnow (2009).

Score-based generative models generate new samples by integrating a time-dependent drift that carries Gaussian noise onto the target distribution. In practice this drift is modeled by a neural network, trained on a loss integrated over time $t$ with a weighting schedule $w(t)$. Along the backward dynamics, and for multi-modal distributions, trajectories commit to modes of the target within a narrow time window, the speciation time. In this work, focusing on high-dimensional data, we decompose the integrated loss into its single-time contributions and analyze each at fixed signal-to-noise ratio $\Lambda(t)$: we show that $\Lambda(t)$ sets the rate at which each feature of a multimodal target - the mode directions and their relative weights - is acquired during training. Crucially, at high $\Lambda(t)$ all mode directions are acquired together, on a single timescale insensitive to their amplitudes, while the relative weights are not learned at all. Only near the speciation time, where $\Lambda(t)$ becomes of order one, do all features become learnable, each on its own timescale: the weights are acquired jointly with the directions, and the directions at rates set by their relative amplitudes. For models trained on time-integrated objectives, the learning dynamics is then governed by how much of the weighting effectively sits near the speciation time, which provides insights on $w(t)$ design choices. These results follow from an exact high-dimensional analysis of the training dynamics of unbalanced and hierarchical Gaussian mixtures. Numerical experiments on image and human genome haplotype generation recover the predicted hierarchy of learning timescales in more complex settings.

On-policy distillation (OPD) trains a student on its own trajectories with dense teacher supervision. Recent work on OPD for multi-turn autonomous agents often treats large teacher-student token-level distributional gaps as promising intervention points, linking larger gaps to a greater need for correction. Yet, our empirical analysis reveals a supervision-benefit mismatch: large gaps can be benign, while small gaps can be outcome-critical. Teacher-student gaps capture differences at the current turn, whereas the benefit of teacher guidance depends on how the current student interacts with the environment afterward. The student may still succeed despite choosing an action that differs from the teacher's, while a teacher-preferred action may lead to a state from which the student cannot complete the task. Local gaps alone are therefore not enough to determine whether teacher guidance benefits the current student. Effective supervision should instead emphasize guidance that the current student can translate into better final task outcomes. Accordingly, we propose Outcome-Guided On-Policy Distillation (OG-OPD), which applies trajectory-relative weighting to teacher supervision and calibrates these weights using final task outcomes from paired student continuations. This calibration selectively strengthens supervision on the student's original trajectories at turns where teacher guidance benefits the current student. Across ALFWorld, ScienceWorld, and WebShop, OG-OPD consistently outperforms baselines under diverse settings. It improves task success rates by 3.6-17.7 percentage points over vanilla OPD and by up to 7.0 percentage points over the strongest baseline.

Human videos offer a scalable source of demonstrations for dexterous robot manipulation. However, existing human-to-simulation-to-robot (Human2Sim2Robot) pipelines rely on predefined procedures that struggle to accommodate diverse object properties and interactions, particularly those involving articulated and deformable objects. We introduce DexAgent, an agentic Human2Sim2Robot framework that converts a single egocentric human video and a task prompt into physically grounded robot trajectories for policy training. It operates through four stages: semantic understanding of human videos, property-based simulation reconstruction, robot trajectory optimization, and robot data generation. At each stage, DexAgent adapts its approach to the task and object properties by selecting suitable skills from its tool library or developing new ones when needed. Property-specific verifiers assess stage outcomes for physical validity and task-specific requirements and provide feedback for refinement, preventing error propagation through the workflow. This adaptive, verification-guided process allows DexAgent to process diverse objects and long-horizon tasks. In the final stage, DexAgent varies object and robot states in simulation to generate diverse robot trajectories from a single human video, then retextures the rendered observations to facilitate sim-to-real transfer. Newly developed skills and verifiers are retained in its tool library, making it self-evolving to accumulate reusable capabilities. This reduces processing time as DexAgent encounters more human videos. Across eleven real-world tasks, policies trained with DexAgent-generated data achieve a 3.5x higher success rate than competing baselines. Project website: https://dexagent1.github.io/.

AI reliability concerns whether an AI system performs its intended function dependably over a stated period and under stated operating conditions, with stated evidence. As these systems become more autonomous, that function includes more than a correct output. Retrieval, memory, tool use, permissions, human oversight, and interactions among systems must operate consistently and safely, and, for generative systems, so must the reasoning process that produces the output. Average benchmark accuracy measures capability; it does not quantify this broader reliability claim. This paper adapts established reliability engineering methods, from failure definitions and operational envelopes to FMEA, accelerated testing, field monitoring, and reliability growth, to AI systems. A four-level diagnostic framework classifies failures as component, operational-loop, agentic-conduct, or network and governance failures. Test, evaluation, verification, and validation (TEVV), sequential monitoring, and FRACAS create and refresh evidence. SMART provides statistical guidance for measurement, analysis, assessment, and test planning; the NIST AI Risk Management Framework provides organizational guidance for governance, evaluation, monitoring, and mitigation. Three cases illustrate the program: adversarial testing of a convolutional neural network, perception-error propagation, and autonomous-vehicle disengagements. Established reliability engineering provides a usable foundation; new measurements and safety guardrails are still needed as these systems are self-evolving.

Large Language Model (LLM)-based agents promise to automate scientific discovery, yet exploring the vast hypothesis space remains costly. Existing principle-evolution methods accelerate this loop, but operate sequentially, which caps exploration breadth and wastes wall-clock time on challenging problems. To address this, we formulate collaborative scientific discovery as evidence transfer between parallel principle-evolution branches. We present COEVOLVE, which realizes this transfer through a coordination core over parallel branches. By integrating value-of-information-gated routing and context-discounted likelihood injection, COEVOLVE enables branches to collaborate through shared measurements while keeping their principle posteriors separate. Across six scientific-discovery tasks under a matched evaluation budget, COEVOLVE attains a mean solution quality of 66.5% versus 57.0% for single-branch principle evolution, with a 1.80x mean wall-clock speedup on the GPT-5.6-Terra backbone; on five auto-research tasks delegated to an autonomous research harness, it is the only arm whose mean stays above the published SOTA anchor on every task. These results establish when evidence sharing accelerates parallel discovery and when transfer safeguards are necessary to limit negative or inert transfers

Neural residuals can improve satellite evapotranspiration (ET) estimates, but selectors must predict when a correction helps and reject unsupported inputs. We evaluate ten-member models on 16,366 flux-tower observations from 151 stations paired with OpenET, across nine rolling years and five spatial folds. At one held-out station, Gain accepted corrections on all 32 physically invalid records: it predicted a mean benefit of 0.83 mm/day, but the corrections increased mean absolute error by 21.6 mm/day versus OpenET. On spatially held-out unit errors, SupportGain reduced station-macro MAE versus Gain by 0.148 mm/day under wind x3.6 (simultaneous 95% interval, 0.070 to 0.226), with 9.3% acceptance versus Gain's 51.8%; on clean inputs, its 0.006 mm/day advantage had an interval that includes zero. These fault analyses are exploratory; none of 40 preplanned temporal comparisons passed Holm correction, while a separate predeclared cropland contrast found 0.041 mm/day lower station-macro MAE with crop-only training (95% interval, 0.009 to 0.079).

Large Language Models (LLMs) perform well on reasoning benchmarks, but it remains unclear whether this reflects genuine contextual reasoning or reliance on facts memorized in their parameters. We investigate this by distinguishing two possibilities: a broad memorization bias, where familiar content improves reasoning performance, and the Strong Parametric Shortcut Hypothesis, where models skip reasoning entirely and recall stored answers. To test these effects, we introduce MemoReason, a human-curated benchmark that pairs factual reasoning tasks with structurally identical \fictitiousterm{} versions where real entities like people, companies, or dates are systematically replaced by \fictitiousterm{} ones of the same type. This \scorerevision{preserves task structure and specified reasoning operations} while varying the familiarity of the context, allowing controlled measurement of how the parametric memory affects reasoning. \revision{Our evaluation of recent LLMs reveals consistent and statistically significant performance drops of up to 15.7% in the fictitious setting, demonstrating a clear memorization bias.} However, a targeted analysis of \revision{questions failed in the fictitious setting} shows that models rarely respond with the corresponding factual answer, indicating that direct parametric shortcuts are not the dominant failure mode. These findings suggest that parametric memory influences reasoning through mechanisms more complex than simple factual recall. MemoReason provides a controlled framework for studying these mechanisms and for extending paired factual-fictitious{} evaluation to broader reasoning settings.