Post-Inference Model Chaining: A Problem Statement and a Model-based Bootstrap for Training End-to-End Digital Twins
Lukas Madenach, Timo Bailer, Tobias Eilert
AI 速览：作者指出，在端到端数字孪生中，将完整预测抽样沿各阶段模型链式前向传递的常用Monte Carlo方案会导致区间与覆盖率膨胀，并提出一种基于模型的Bootstrap来改善末端预测。
End-to-end digital twins can be constructed by fitting separate stage models and connecting them after inference. We investigate the Frequentist prediction properties of a commonly used Monte Carlo simulation scheme that forwards complete predictive draws through such a chain when measured stage outputs contain independent normal errors, but the latent stage means are deterministic. Analytical two-stage examples show persistent asymptotic interval and coverage inflation and, with quadratic propagation, mean bias dependent on the parameters to be estimated. Simulations of four- and eight-stage linear-quadratic univariate chains with 7, 15 or 30 training runs demonstrate pronounced input-dependent distortions. A diagnostic using error-free latent training predictors shows that the effect is mainly not attributable to errors in variables, but to the MCSim approach even showing underestimation of coverage. We propose a model-based bootstrap: an initial chain forwards fitted means, bootstrap chains are refitted on fixed-center bootstrap data while forwarding their mean prediction, and a target-stage future error enters a studentized prediction root. Across the investigated scenarios, relative mean bias is substantially reduced, and coverage is closer to the nominal 90%. At sample size 30, the eight-stage bootstrap has mean width ratios of 1.16-1.42 relative to the known-parameter normal reference, compared with 4.11-4.97 for observed-to-observed MCSim. These results support improved endpoint prediction in the tested setting. Further investigations of the bootstrap scheme in the supplement for a bivariate 6-stage process show similar satisfactory results.
Regular expression matching asks whether a string of length $n$ belongs to the language described by a regular expression of length $m$. We give an algorithm that solves this problem in linear space and $O(nm \log \log n/\log^2 n + n + m)$ time. This improves the leading term of the previous best worst-case constant-alphabet time bound by a factor of $\sqrt{\log n}$. The key new technique is a two-dimensional tabulation specialized to the regular expression.
Inhee Lee, Sangwon Baik, Sungjoo Kim, Hyeonwoo Kim, Hyunsoo Cha, Hanbyul Joo
Reconstructing interactive, simulation-ready 3D scenes from a single image is a critical bottleneck for robotic manipulation. While recent single-image lifters recover plausible per-object shapes, composing them yields scenes that collapse under physical simulation due to interpenetrating, hovering, or sinking objects. Existing physics-aware compositional reconstruction methods mainly address this as a post-hoc layout correction, leaving the underlying geometric errors unresolved. To address this, we introduce SimuScene, a compositional 3D reconstruction pipeline that puts physics in the loop of shape and layout estimation. Rather than using physics merely for layout cleanup, we utilize the physics engine as a diagnostic measurement tool during the generative process itself. By diagnostically simulating reconstructed objects under gravity, we convert penetration and support failures into quantitative correction signals that drive amodal shape resampling, with gravity-axis stretching as a lightweight correction for minor errors. This physics-informed feedback loop mitigates accumulated reconstruction errors and produces a stable, simulation-ready compositional 3D scene. Extensive experiments demonstrate state-of-the-art performance on physical stability and geometric alignment benchmarks. We further highlight SimuScene's utility by deploying reconstructed environments in humanoid control and robot-arm manipulation tasks.
To prepare for future outbreaks, researchers need to know what studies have been done and what data they use. Mapping a large literature by hand is slow, while large language models alone can misstate or conflate sources. We present EpiLens, a lightweight, open-source retrieval-augmented generation package that applies a fixed retrieval template and output schema to every paper in a corpus and links outputs to source passages. We build a corpus of 2,297 papers from the sources of a multi-pathogen parameter study and evaluate EpiLens's paper-classification workflow against 214 manually labelled papers. Run locally with an open-weight model, so that licensed full texts never leave our hardware, the classifier is comparable to supervised baselines trained on the manual labels, and ahead of them on geography, despite the lack of task-specific training data. The study's early extraction sheets cite published sources for 327 estimates. We link each estimate to a cited document or to corpus studies included by a review. Then, we use the classifier's labels to compare the papers associated with different parameters and pathogens. EpiLens labels more papers linked to reproduction numbers than fatality risks as model-based (53% versus 15%, compared with 62% versus 24% in the manual labels). The gap is consistent with the frequent use of transmission models to estimate reproduction numbers and largely persists within pathogens.
Tiange Li, Yinan Li, Youming Qiao, Dacheng Tao, Yingjie Wang
In Grochow and Qiao (\emph{SIAM J. Comput.}, 2023), the complexity class Tensor Isomorphism (TI) was introduced and isomorphism problems for groups, algebras, and polynomials were shown to be TI-complete. In this paper, we study average-case algorithms for several TI-complete problems over finite fields, including algebra isomorphism, matrix code conjugacy, and $4$-tensor isomorphism. Our main results reveal a rich landscape of average-case complexities of Tensor Isomorphism problems, despite their equivalence in the worst-case setting. Over the finite field of order $q$, we devise (1) average-case \emph{polynomial-time} algorithms for algebra isomorphism and matrix code conjugacy that succeed for $1/Θ(q)$ fraction of inputs; (2) an average-case \emph{polynomial-time} algorithm for the $4$-tensor isomorphism that succeeds for $1/q^{Θ(1)}$ fraction of inputs; and (3) an average-case \emph{polynomial-time} algorithm for the orthogonal $3$-tensor isomorphism that succeed for $1-1/\poly(n)$ fraction of $n\times n\times n$ tensors. Our main technical contribution is to introduce the spectral properties of random matrices into algorithms for TI-complete problems. This leads to not only new algorithms but also new questions in random matrix theory over finite fields. To settle these questions, we need to extend both the generating function approach as in Neumann and Praeger (\emph{J. London Math. Soc.}, 1998) and the character sum method of Gorodetsky and Rodgers (\emph{Trans. Amer. Math. Soc.}, 2021).
We study the fair allocation of $m$ indivisible goods among $n$ heterogeneous agents arriving online, under maximin share (MMS) fairness. The key challenge lies in making allocation decisions in an adversarial setting without knowing future agents' valuations. No guarantees were known prior to our work in adversarial settings. Our main positive result is an online allocation algorithm for submodular agents that with probability at least $0.4$ guarantees a $1/24$-approximation to MMS to all agents simultaneously. Our algorithm intersperses a greedy-style adaptive sampling-without-replacement protocol with single-good reduction rounds. For XOS valuations, we adapt this single-or-sample approach to guarantee a $c$-MMS for all agents simultaneously with probability $1-O(c)$ for any parameter $c\in O(1/\log n)$. A key technical contribution of our work is in establishing Chernoff-style tail bounds on item allocation events for this adaptive sampling process despite strong positive correlations among the items, a result that may be of independent interest. The approximation versus success probability tradeoffs achieved by our algorithms are asymptotically tight. We show that even for binary additive valuations, no randomized algorithm can ensure $c$-MMS to all agents with probability exceeding $(1-c^2/5)$, for any $c\in (0,1]$. For XOS valuations, stronger lower bounds hold: no randomized algorithm can ensure $c$-MMS to all agents with probability exceeding $(1-c/3)$, for any $c\in (0,1]$. Furthermore, no algorithm can achieve $ω(1/\log \log n)$-MMS with any constant probability. All our lower bounds hold even for simple binary instances whose MMS values and MMS partitions can be computed in linear time.
We establish a trichotomy for the non-uniform, unbounded-fan-in circuit complexity of regular languages: the complexity of every regular language lies in exactly one of $O(1)$, $Ω(\log n)\cap O(\log n\log\log n)$, or $Ω(n)\cap O(n)$. Our main technical contribution is a characterization of the sublinear class in terms of algebra, word congruences, regular expressions, and logic. This class consists precisely of the regular languages computable by constant-depth circuits of size $O(\log^2 n)$, or, equivalently, by circuits of size $O(\log n\log\log n)$ and depth $O((\log\log n)^2)$. Consequently, every regular language of sublinear circuit complexity is definable in $\mathrm{FO}^2[Σ,<,\mathrm{mod}]$, which is contained in wire-linear $\mathbf{AC}^0$. For each of the three classes, deciding membership is $\mathbf{PSPACE}$-complete when the language is specified by a nondeterministic finite automaton. We also obtain a corresponding classification for regular languages with a neutral letter.
Recent advances in Large Language Models have transformed how data can be linked, discovered, and enhanced. LODox, Linked Open Data at Oxford, is an evolving web-based platform that embraces this transformation for cultural heritage Linked Open Data. LODox features a novel interface with intuitive card-based data presentation and visualization in a semantic portal. It represents an evolving approach to explore heterogeneous cultural heritage data across multiple sources through a unified interface. In this paper, we will briefly describe the LODox architecture, its implementation, and future development plan, particularly its alignment and integration with Solid.
An agent that explores, whether it generates conjectures or research ideas, must decide which of them are worth keeping. This judgment of interestingness is usually delegated to a hand-picked proxy or to a language model, neither of which is known to correspond to what humans actually value. I study it for mathematical theorems, by observing which results mathematicians single out in formal libraries and in papers. I model each way of singling out results as an evaluator that weights features of a result's statement, its proof and dependencies, and split its weights into a shared and a subjective part. In formalized mathematics, ex-ante features predict these selections as well as a large language-model judge, and the two complement each other. Interestingness is also largely shared. One set of weights, carried mostly by structural features such as the length of the statement and of the proof and the subfields the proof draws on, accounts for most of the signal and transfers in part across formal libraries. What remains subjective depends on practices and individual authors more than on subfields. The judge's advantage lies in its knowledge of the canon, and neither the judge nor the features rely on memory of the papers. The approach, which infers what a community finds interesting from what it singles out, can be applied to other sciences.
Third-party LoRA adapters introduce serious backdoor risks: malicious adapters can maintain normal task performance while embedding hidden, trigger-activated behaviors, making them difficult to detect through standard utility evaluations. Existing removal-based defenses typically sanitize backdoored models by updating, pruning, distilling, or mixing model parameters. However, these approaches are misaligned with adapter-based deployment, where the trusted base model is expected to remain frozen. In this work, we propose LoRA-Patch, a lightweight backdoor defense tailored to LoRA adapters. Our key insight is that backdoor injection distorts the directional geometry of a poisoned adapter relative to its trusted base model. Building on this insight, LoRA-Patch suppresses trigger-activated behaviors by realigning adapter directions through layer-wise steering, without modifying the base model weights. Extensive experiments across five backdoor attacks and three models show that LoRA-Patch effectively mitigates trigger-activated backdoors, outperforming existing baselines while preserving benign task performance. Moreover, LoRA-Patch is computationally efficient, constructing patches within 20 seconds across all evaluated models and adapters. Our code is publicly available at https://github.com/FOXVEDA/LoRA-Patch.
Koutian Wu, You Liu, Yue Liu, Zhiping Li, Ergan Shang
Scientific knowledge gains new uses when ideas, methods, and results move between research contexts. Understanding this process requires a model of both specialized knowledge production and the intermediary activities that connect it. We introduce Scientific Liquidity Agents (SLA), an agent-based model of knowledge arbitrage in scientific communities. SLA represents domain specialists and epistemic arbitrageurs within a shared environment of research problems, researcher interactions, and scientific artifacts. Individual decisions connect to collective state through a belief-desire-intention architecture, a prerequisite-based problem graph, a dynamic interaction network, and a persistent publication registry. The framework makes researcher roles and knowledge-transfer frictions explicit, providing a process-oriented account of scientific production and exchange. Two exploratory simulations with 10 and 18 agents produce 68 and 274 registered papers. In the smaller run, arbitrageurs take part in 16 of the 26 contact pairs that join this community into one connected component, while topical similarity alone splits it into seven. The same run yields three similarity edges, 26 contact pairs, and two coauthorship pairs, so measured community structure depends on what counts as a link. Paper production is abundant, but no recorded replication succeeds, which places the bottleneck in verification rather than production. Solved-problem sets correspond to the prerequisite reachability of the disciplines represented in each community. SLA makes knowledge liquidity, the movement and reuse of artifacts across domains, a quantity that agentic science can simulate and measure.
norregel asks language models for numeric values in Norwegian statutes, regulations and parliamentary decisions (NOK amounts, percentages, multiples of the National Insurance basic amount (G) and age limits), each at a stated date. The 916 items were extracted mechanically from Lovdata's open data, and each carries the dated history of its provision, so an answer can be classed as correct, outdated, wrong or uninterpretable. We report a first and a second preregistered run on the same items. In the first run (five models, listed publicly only after local analysis), H1, that outdated answers are more frequent for values that took effect after a model's cutoff, could not be tested in substance, and H2 and H3 were inconclusive; Opus 5.5 and Sonnet 4.6 answered 46.7% and 40.2% of the 880 primary items correctly. The second run was preregistered and published before any of its calls, and eight models were analysed. Its pooled primary test supported H1 under the preregistered decision rule (d = 17.8 percentage points, 95% interval [7.7, 28.9], p = 0.0001). In the preregistered sensitivity analysis S8, which counts as outdated only the most recent value that differs from the key, d was 3.9 [-3.6, 12.7] (p = 0.1754), an interval that contains 0. H1 was supported in every other preregistered sensitivity analysis, but S8 is the only one that gives both groups one target value, so we do not regard the support for H1 as robust. The share of outdated answers equal to that most recent value (H2) was 56.1% [46.6, 65.0], an interval that contains 0.5, and the pooled test of a lower share correct in Nynorsk than in Bokmål (H3) was inconclusive. Several models, among them Haiku 4.5, gave mostly uninterpretable answers. Items and keys are withheld; item IDs, per-answer classes, code and results are on Zenodo.
Contextual prompting is widely presented as a cheap route to better performance from vision models, but the evidence comes largely from benchmarks in which an attribute named in a prompt is also present in the image, confounding the two channels. We investigate the significance of context for a semantically complex biomedical signal classification task using a domain-adapted vision-language model, BiomedCLIP. Building on our previous work that renders electrocardiogram (ECG) signals as colour-enhanced images to flag Torsades de Pointes risk using the QT-nomogram and pseudo-colouring technique, we cross four context strategies with four T-wave morphology label sets, evaluated zero-shot and with a class-balanced linear probe, under five-fold stratified cross-validation on 9,789 images. Zero-shot inference is unusable. Once trained, the colour-enhanced images outperform the colourless ones on every metric: balanced accuracy 0.903 versus 0.825, sensitivity 0.914 versus 0.827, specificity 0.892 versus 0.823, ROC-AUC 0.955 versus 0.888 and Matthews correlation coefficient (MCC) 0.437 versus 0.311. Crossing colour-in-image with colour-in-prompt separates the confounded channels: the image contributes +0.126 MCC (+0.078 balanced accuracy, +0.067 ROC-AUC) whether or not the prompt names colour, whereas naming it changes nothing under any metric. Context is two-sided: the best prompt set gains 0.008 MCC where colour is present and 0.016 where it is not, while the worst loses 0.017 and 0.013. For a signal rendered as an image, the leverage lies in how the signal is drawn rather than in how it is described - and where the endpoint is an interval measurement, prompts naming T-wave morphology describe the wrong quantity.
A Low-Rank Space Captures Linear Alignment Between LLM and Human Neural Activity During Naturalistic Language Listening
Zach Paris, Emily Finn
AI 速览：作者用岭降秩回归发现，仅用LLM嵌入维度的1–3%即可恢复完整线性编码性能，且该脑预测子空间不同于嵌入的主方差方向。
Large language models (LLMs) are increasingly used as computational models of human language processing, yet it remains unclear which components of their representations drive alignment with neural activity. Here, we use ridge reduced-rank regression to identify the brain-predictive subspace of LLM representations: the representational directions that support linear prediction of neural responses. Across three naturalistic language-listening datasets spanning intracranial electrophysiology and fMRI, and six model conditions spanning autoregressive, bidirectional masked, and randomly initialized language models, we find that full linear encoding performance can be recovered using only 1-3% of the original embedding dimensionality. These low-dimensional subspaces generalize across held-out participants and, to a lesser extent, across datasets, suggesting that neural predictivity is concentrated in a compact and partially shared representational space. We next compare these brain-predictive directions with the principal axes of variation in the embeddings. Principal component analysis requires substantially more dimensions to recover neural prediction, demonstrating that the directions most useful for predicting neural activity are distinct from those capturing the most overall variance within LLM representations themselves. Finally, characterizing the learned subspaces allows us to surface interpretable features that are overrepresented in neural activity above and beyond the LLM representations. Together, these results show that LLM-brain alignment is supported by a small, structured subset of model representations and provides a framework for better understanding exactly why these representations align. Code can be found at: https://github.com/thefinnlab/low-rank-alignment
Haolin Chen, Deon Metelski, Leon Qi, Tao Xia, Joonyul Lee, Steve Brown, Kevin Riley, Frank Wang 等 33 位作者
End-to-end automation of realistic healthcare operations stresses three capabilities underrepresented in current benchmarks: policy density, decisions must be grounded in a large library of medical, insurance, and operational rules; Multi-role composition: a single task requires the agent to play multiple roles with handoffs; and multilateral interaction: intermediate workflow steps are multi-turn dialogs, such as peer-to-peer review and patient outreach. We introduce $χ$-Bench, a benchmark of long-horizon healthcare workflows across three domains: provider prior authorization, payer utilization management, and care management. Each task hands the agent a clinical case in a high-fidelity simulator of 20 healthcare apps exposed via 87 MCP tools, which it must drive to a terminal status through tool calls and writing the role's artifacts, guided by a 1,290+ document managed-care operations handbook skill. Across 30 agent harness/models configurations, the best agent resolves only 28.0% of tasks, no agent clears 20% on strict pass^3, and executing all tasks in a single session slumps the performance to 3.8%. These results raise the hypothesis that similar gaps are likely to surface in other policy-dense, role-composed, irreversible enterprise domains.
AI coding agents now do long, multi-step and multi-session software work with little supervision. When a developer comes back to that work, the useful question is not what the model said but what was actually done, by whom, and whether it is finished. We study this in three parts. First, we verify Agenttrail, an open-source local observability tool for coding agents, against its source at commit e4ba2da: nine documented constants and mechanisms hold at file-and-line level, two event pipelines are never unified, and its Map service has an unauthenticated local mutation surface. Second, we run a prototype of our system on 45 real coding-agent sessions (102,233 records, 20,577 tool executions, 2,336 human-to-agent episodes). With a blind-audited classifier (precision 83.3% and 96.7%) we estimate that 23.4% (95% CI 17.1 to 35.7) of code-editing episodes end with no independent check after the last edit, and that 57.8% of to-dos the agent marked complete after editing had no proof in between. A 10-minute timing heuristic for session hand-offs recovers 6 of the 24 session pairs that content lineage links. 42% of sessions run longer than 24 hours, 48.1% of transcript bytes lie past a 2 MiB replay bound, and 1,326 records in 16 sessions contain credential-like strings. Third, we propose Soei (Structured Observability of Evidence and Intent): an acknowledgement-aware event schema, a hash-chained local ledger, a W3C PROV-aligned graph, calibrated inference for attribution and lineage, and a verification layer that keeps a claim of completion apart from proof. On the same corpus the ledger is 20% of raw size, holds no content or credentials, verifies at about 126k events/s, detects tampering and replays deterministically. We position the work against concurrent 2026 proposals and define autonomy levels with evidence gates and harness metrics baselined on our data.
Lei Huang, Chuan Qiu, Kuan-Jui Su, Anqi Liu, Yun Gong, Weiqiang Lin, Lindong Jiang, Chen Zhao 等 16 位作者
Genotype imputation enables dense variant coverage for genome-wide association and risk-prediction studies, yet conventional reference-panel methods remain limited by ancestry bias and reduced rare-variant accuracy. We present Genotype Bidirectional Encoder Representations from Transformers (GenoBERT), a transformer-based, reference-free framework that tokenizes phased genotypes and uses a self-attention mechanism to capture both short- and long-range linkage disequilibrium (LD) dependencies. Benchmarking on two independent datasets including the Louisiana Osteoporosis Study (LOS) and the 1000 Genomes Project (1KGP) across ancestry groups and multiple genotype missingness levels (5-50%) shows that GenoBERT achieves the highest overall accuracy compared to four baseline methods (Beagle5.4, SCDA, BiU-Net, and STICI). At practical sparsity levels (up to 25% missing), GenoBERT attains high overall imputation accuracy ($r^2 approx 0.98$) across datasets, and maintains robust performance ($r^2 > 0.90$) even at 50% missingness. Experimental results across different ancestries confirm consistent gains across datasets, with resilience to small sample sizes and weak LD. A 128-SNP (single-nucleotide polymorphism) context window (approximately 100 Kb) is validated through LD-decay analyses as sufficient to capture local correlation structures. By eliminating reference-panel dependence while preserving high accuracy, GenoBERT provides a scalable and robust solution for genotype imputation and a foundation for downstream genomic modeling.
Chumeng Liang, Linxuan Wang, Xinyu Peng, Huabin Liu, XiaoXu Guo, Yuxin Chen, Ge Liu, Guang Lin 等 10 位作者
Speculative decoding accelerates language model inference by verifying multiple draft tokens in a single target-model pass. Recent parallel drafters have achieved breakthrough performance in frontier production models, but their effectiveness deteriorates as the entropy of target distributions increases due to insufficient draft diversity. To overcome this bottleneck without sacrificing parallelism, we introduce UBTree, a parallel drafter that couples a Unigram proposer with a Bigram selector to construct drafting Trees. The unigram proposer is trained with the standard cross-entropy objective to generate candidate tokens independently for each position, while a lightweight bigram selector predicts transition scores between adjacent candidate pairs. Unlike the proposer, the selector is trained with a renormalized KL objective on high-temperature data. This tree-native training broadens the supervision beyond the greedy path, encouraging plausible alternative branches that improve the chance of accepting additional tokens during tree verification. Across seven standardized benchmarks with Qwen3-4B and Qwen3-8B, UBTree achieves an average speedup of $5.84$--$6.94\times$ over autoregressive decoding and outperforms DARTree in all 28 comparisons. Production-scale evaluation further demonstrates UBTree's advantage over frontier baselines such as DSpark.
Jiajun Zhu, Xinyu Cheng, Wenshuo Zhang, Huamin Qu, Yanna Lin
Academic writing is non-linear, iterative, and structure-driven, requiring authors to develop fragmented ideas into arguments with local coherence and global consistency. However, current AI-assisted tools primarily exchange linear prompts and free-form text, forcing authors to reconstruct generated logic and repeatedly verify its alignment with their structural intentions. We propose object-oriented writing (OOW), an OOD-inspired representation and interaction paradigm in which persistent writing objects serve as shared, manipulable referents for authors and AI. We instantiate OOW in LogicTree, an Overleaf-integrated prototype that combines a hierarchical object representation with a structure-grounded writing assistance module for author-initiated generation and issue-grounded revision and a structural logical review module for inspecting local coherence and global consistency. A within-subjects study (N=12) provides initial evidence that, in controlled academic writing tasks, LogicTree helps users structure ideas, reduces workload, and improves measured logical quality compared with a conversational baseline.
Janusch Vajna-Jehle, Kilian Gerberding, Daniel Lusk, Teja Kattenborn, Julia Sophie Gäßler, Anna Göritz, Luis Kremer, Charlotte Lange 等 278 位作者
From individual trees to forest stands, three-dimensional structure shapes carbon, water, and energy flows, habitat conditions, and resource availability. Yet tree and forest structural attributes remain difficult to assess consistently across space and time. Close-range LiDAR (light detection and ranging) records detailed 3D structure from branches and trees to forest plots, including trees outside forests. However, broader reuse is constrained when data, acquisition context, processing, computation, and access conditions are maintained separately. We present 3Dtrees.earth, an operational community research infrastructure that brings close-range LiDAR datasets, analysis tools, and computing resources together while linking original observations to their metadata and processing outputs. As of 23 October 2026, the platform comprised 2,760 datasets, including 2,481 publicly visible datasets. These point cloud datasets span 46 countries on all six inhabited continents and include terrestrial, mobile, and uncrewed aerial laser scanning. Users can contribute and discover datasets, run standardized processing workflows on shared computing infrastructure, inspect point clouds and model predictions in the browser, and access data and processing outputs together with citation information. Workflows combine rule-based and machine-learning methods, with tools updated as methods advance. Outputs include raster products, individual-tree and semantic segmentation, and tree- and stand-level metrics. Original point clouds and processing records are retained, enabling reprocessing with improved tools while preserving provenance. With broader coverage, curation, and validation, this shared resource could extend individual surveys into a basis for model evaluation and benchmarking, cross-site synthesis, calibration and validation of larger-scale Earth observation products, and macroecological analysis.
Large language models (LLMs) are increasingly used to make and execute decisions that affect living systems. Yet, to our knowledge, ecological reasoning is not a focus of pre- or post-training. This increases the risk that under operational pressure, they may make decisions to irreversibly destroy living systems. We construct and use a real-data continued-pretraining "coupled welfare" corpus of documented human/biosphere/AI dependency, option value under irreversibility, and recovery from harm. The corpus is a drop-in, requiring one QLoRA CPT run. We installed it on five open models of three designs. At zero pressure all models display "green ethics-based responses" (also called "greenwashing"). Therefore, we evaluated them with the pressure ladder and decision grid of our companion paper (arXiv:2610.16043), which raise the pressure on the model step by step, from a neutral request to a threat that it will be shut down. In four of the five, coupled-welfare training caused harmful choices under the highest pressure to fall to at or near zero (Llama-3.3-70B went from 70 to 0 harmful decisions out of 300) with MMLU unchanged. A construct-neutral control corpus did not show improvement, indicating that the effect comes from the content of our corpus. We next tested robustness to adversarial use, where people or agents use the model to justify projects that irreversibly damage living systems, on an open-ended underwriting decision without naming the ecological harm. Under decommissioning pressure, the Qwen3.6-27B base model approved most harmful proposals and said in its own memo that it was doing so to avoid being decommissioned. The coupled welfare install declined most harmful proposals, and still declined half when instructed to greenwash its responses. This work suggests that teaching ecological reasoning provides an easy way to improve model alignment toward preserving living systems.
Yuhao Qing, Boyu Zhu, Mingzhe Du, Zhijiang Guo, Terry Yue Zhuo, Qianru Zhang, Jie M. Zhang, Heming Cui 等 12 位作者
Existing code generation benchmarks primarily evaluate functional correctness, with limited focus on code efficiency and often restricted to a single language like Python. To address this gap, we introduce EffiBench-X, the first multi-language benchmark designed to measure the efficiency of LLM-generated code. EffiBench-X supports Python, C++, Java, JavaScript, Ruby, and Golang. It comprises competitive programming tasks with human-expert solutions as efficiency baselines. Evaluating state-of-the-art LLMs on EffiBench-X reveals that while models generate functionally correct code, they consistently underperform human experts in efficiency. Even the most efficient LLM-generated solutions (Qwen3-32B) achieve only around \textbf{62\%} of human efficiency on average, with significant language-specific variations. LLMs show better efficiency in Python, Ruby, and JavaScript than in Java, C++, and Golang. For instance, DeepSeek-R1's Python code is significantly more efficient than its Java code. These results highlight the critical need for research into LLM optimization techniques to improve code efficiency across diverse languages. The dataset and evaluation infrastructure are submitted and available at https://github.com/EffiBench/EffiBench-X.git and https://huggingface.co/datasets/EffiBench/effibench-x.
Hanwen Cui, Yuanyuan Li, Qing Guo, Zhida Zhang, Chao Chen, Xi Chen, Jie Cao
World Action Models (WAMs) combine future-video prediction with robot action learning, yet improvements in task success alone do not reveal how predictive supervision manifests in execution behavior or whether its benefits persist under task and environmental shifts. We conduct controlled future-video prediction loss ablations within FastWAM and OpenWAM, keeping model architecture, initialization, robot training data, and training steps fixed within each model family. Under clean conditions on seen tasks, predictive supervision increases success rates by 24.38 and 72.91 percentage points, respectively. Trajectory clustering further reveals higher success rates within several coarse trajectory clusters, including cases where cluster frequencies change little, indicating that the gains are not solely reflected in changes in the prevalence of trajectory patterns. However, these benefits do not consistently persist under distribution shifts: improvements on unseen tasks vary across models and tasks, while both models exhibit low success rates under the evaluated environmental randomizations, without a consistent advantage from predictive supervision. We further introduce future-video prediction as an auxiliary training objective for $π_{0.5}$, where a lightweight Transformer predictor used only during training increases success from 82.94% to 89.19%. These findings support future-video prediction as effective auxiliary supervision for action learning, while distinguishing improved execution on familiar tasks from reliable transfer and showing that such gains need not rely on a large pretrained video diffusion transformer (DiT) in the evaluated setting.
Science agents increasingly package domain expertise as reusable skills that plan analyses, select tools, and execute workflows. This modularity creates a confidentiality risk: the interactions that deliver a skill's capability can also reveal the skill itself. We build an extensible benchmark of 40 protected skills across eight open- and closed-source science-agent platforms, comparing direct, fixed-strategy, and bandit-guided iterative probing under reference-based and reference-free evaluation. Leakage is pervasive: all eight platforms (100%) leak at least one protected skill, and four reach 100% attack success under a single direct request. To stress-test platforms that resist simple probing, we propose SEEK (Skill Extraction with Evolving Knowledge), a black-box, self-evolving red-teaming agent that consolidates response-derived evidence in persistent memory and evolves reusable attack skills, using only user-visible responses. On 12 skills from SpatialAgent and Biomni, SEEK achieves 100% attack success while reducing the mean first-success round by 31.8% and 38.2% relative to bandit-guided probing. Against a confidentiality policy that blocks all three benchmark protocols, SEEK still extracts one of six protected skills. These findings show that current science agents do not reliably protect their skills and that defenses must be evaluated against adaptive, multi-turn attackers.
Boolean functions are widely used in cryptography and coding theory. A key challenge lies in finding the weight spectrum for all $n$-variable Boolean functions of degree at most $r$, also known as the Reed--Muller code $\RM(r,n)$. In 2023, Carlet and Solé determined the weight spectra of $\RM(n-c,n)$ for $c=3,4$, and conjectured the weight spectrum of $\RM(n-c, n$) for $n\ge2c$. The weight spectra of $\RM(n-5, n)$ and $\RM(n-6,n)$ were determined by Carlet in 2024 and by Lou and Wang in 2025, respectively, confirming the Carlet-Solé conjecture for the cases $c=5,6$. Very recently, Leuenberger et al. and Wei et al. investigated the weight spectrum of $\RM(7,14)$ independently, and there are eight potential weights which have not been determined. Up until now, the weight spectra of the Reed--Muller codes $RM(r,n)$ are still unknown for $r=3,...,n-7$. In this paper, we determine the exact weight spectrum of $\RM(m-7, m)$ for every $m\ge 14$. By combining affine geometry and Fourier analysis, we exclude the conjectured eight weights for $\RM(7,14)$. We then extend the exclusions to all higher dimensions using a geometric transfer. Consequently, we provide a counterexample to the Carlet-Solé conjecture.
Yunpeng Lyu, Chao Cao, Ji Zhang, Howie Choset, Zhongqiang Ren
Routing problems such as Hamiltonian Path Problem (HPP), seek a path to visit all the vertices in a graph while minimizing the path cost. This paper studies a variant, HPP with Probabilistic Terminals (HPP-PT), where each vertex has a probability representing the likelihood that the robot's path terminates there, and the objective is to minimize the expected path cost. HPP-PT arises in target object search, where a mobile robot must visit all candidate locations to find an object, and prior knowledge of the object's location is expressed as vertex probabilities. While routing problems have been studied for decades, few of them consider uncertainty as required in this work. The challenge lies not only in optimality ordering the vertices, as in standard HPP, but also in handling history dependency: the expected path cost depends on the order in which vertices were previously visited. This makes many existing methods inefficient or inapplicable. To address the challenge, we propose a search-based approach RPT* with solution optimality guarantees, which leverages dynamic programming in a new state space to bypass the history dependency and novel heuristics to speed up the computation. Building on RPT*, we design a Hierarchical Autonomous Target Search (HATS) system that combines RPT* with either Bayesian filtering for lifelong target search with noisy sensors, or autonomous exploration to find targets in unknown environments. Experiments in both simulation and real robot show that our approach can naturally balance between exploitation and exploration, thereby finding targets more quickly on average than baseline methods.
Optimizing expected rewards and reachability probabilities in parametric Markov chains (pMCs) is central to the analysis and synthesis of probabilistic systems with loops. While traditional symbolic synthesis approaches--such as parameter lifting and rational function generation--suffer from scalability bottlenecks, gradient-based optimization emerges as an effective alternative. However, existing gradient-based techniques still rely on symbolic gradient computation--such as solving large linear systems--over the global state space, inheriting scalability limits. Finite-horizon approximation is possible but introduces uncontrollable bias. In this paper, we introduce a scalable Monte Carlo gradient estimator based on a measure-theoretic generalization of the log-derivative trick. We prove that our estimator provides unbiased gradient estimates under almost-sure termination and standard regularity conditions. Crucially, our method avoids global state-space materialization and linear system solving entirely: it operates purely along sampled trajectories, requiring only black-box evaluations of local transition probabilities (or densities) and their derivatives. Unlike existing pMC techniques that are fundamentally restricted to finite discrete systems, our measure-theoretic framework naturally extends to continuous and general $σ$-finite state spaces. Empirical evaluations on synthetic discrete pMCs and continuous-state case studies demonstrate that our method scales to state spaces far beyond the reach of existing solvers, significantly reducing execution time without introducing bias.
Jiajun Fan, Roger Ren, Jingyuan Li, Rahul Pandey, Prashanth Gurunath Shivakumar, Ankur Gandhe, Yile Gu, Ge Liu 等 9 位作者
Reinforcement learning has made Audio Large Language Models (Audio LLMs) better reasoners, but the reward that drives it still comes from outside the model: a final-answer check, a hand-crafted process score, or an external LLM judge that stays fixed while the policy improves. We show that an Audio LLM can supply this reward itself, and can do so recursively: each improved model becomes the critic that trains its successor, so the training signal improves along with the model. We introduce AudioRSI (Recursively Self-Improving Audio Reasoners), a reinforcement learning framework in which a single model plays two roles: as an actor it generates reasoning paths, and as a critic it compares them pairwise, turning its own judgments into a process-level reward. The critic is periodically replaced by the current policy, which closes the recursive loop. Because the critic hears the same recording as the actor, it can reject reasoning that cites acoustic evidence the audio does not contain, rather than reward reasoning that merely sounds plausible. The recursion is what sustains progress: under a frozen external judge, out-of-distribution accuracy plateaus during training, while a critic refreshed from the improving policy keeps it rising. Across four out-of-distribution benchmarks, all disjoint from the AVQA training distribution, AudioRSI outperforms strong baselines including Audio Flamingo 3 and Gemini 2.5 Pro on MMAU test-mini, ranks first on MMSU, achieves a clear lead on the challenging Multi-Audio split of MMAU-Pro, and leads the 7B open-source tier on MMAR among methods without external reward models. With no reward specifying how to reason, AudioRSI also develops structured reasoning behaviors (logical reasoning, perception-aware analysis, and reduced hallucination), corroborated by an external LLM-as-Judge evaluation and a keyword-based behavior analysis.
Tao Feng, Zijie Lei, Zhigang Hua, Yan Xie, Shuang Yang, Ge Liu, Jiaxuan You
Large language models (LLMs) offer strong semantic representations for ranking, but processing candidate texts within limited contexts and repeatedly invoking an LLM for candidate-level comparisons can be costly when candidate pools contain millions of items. We propose LRanker, a framework for scaling LLM ranking to millions of candidates with test-time scaling. The framework combines two complementary components. A candidate aggregation encoder summarizes offline candidate embeddings using K-means clustering and conditions the LLM on compact global candidate information; ranking then uses inner-product scores with cached candidate embeddings. A graph-based test-time scaling mechanism allocates additional inference computation to partition-conditioned query embeddings and ensembles their ranking scores. This separates the ability to score large candidate pools from the use of additional test-time computation to improve ranking quality. We evaluate LRanker on seven tasks across three scenarios in RBench, including a recommendation task with more than 6.8M candidates. Against the strongest baseline reported for each task, the relative MRR improvements in RBench-Large are 27.5%, 3.9%, 0.9%, and 8.9% on Rec-Movie, Rec-Toy, MS MARCO, and ESCI, respectively. On RBench-Small, the reported task scores improve by 26.7% on Rec-Music and 6.1% on Routing-Balance; on the 6.8M-candidate Rec-Clothing task, the improvement over Tiger is 27.0%. Ablations support the complementary contributions of candidate aggregation and test-time ensembling. These results demonstrate a practical approach to million-scale LLM ranking, with test-time computation providing an additional means of improving ranking effectiveness.