arXiv:2509.19102v2 Announce Type: replace
Abstract: General-purpose robotic skills from end-to-end demonstrations often leads to task-specific policies that fail to generalize beyond the training distribution. Therefore, we introduce FunCanon, a framework that converts long-horizon manipulation tasks into sequences of action chunks, each defined by an actor, verb, and object. These chunks focus policy learning on the actions themselves, rather than isolated tasks, enabling compositionality and reuse. To make policies pose-aware and category-general, we perform functional object canonicalization for functional alignment and automatic manipulation trajectory transfer, mapping objects into shared functional frames using affordance cues from large vision language models. An object centric and action centric diffusion policy FuncDiffuser trained on this aligned data naturally respects object affordances and poses, simplifying learning and improving generalization ability. Experiments on simulated and real-world benchmarks demonstrate category-level generalization, cross-task behavior reuse, and robust sim2real deployment, showing that functional canonicalization provides a strong inductive bias for scalable imitation learning in complex manipulation domains. Details of the demo and supplemental material are available on our project website https://sites.google.com/view/funcanon.
Science Journals
arXiv:2509.19590v2 Announce Type: replace
Abstract: Evaluations of generative models are now ubiquitous, and their outcomes critically shape public and scientific expectations of AI's capabilities. Yet skepticism about their reliability continues to grow. How can we know that a reported accuracy genuinely reflects a model's underlying performance? Although benchmark results are often presented as direct measurements of capability, in practice they are inferences: treating a score as evidence of capability already presupposes a theory of what it means to be capable at a task.
We argue that AI evaluations should instead be framed as inference tasks grounded on an explicit theory of capability. While this perspective is standard in fields like psychometrics, it remains underdeveloped in AI evaluation, where core assumptions are often left implicit. As a proof-of-concept, we empirically show that reported performance can depend strongly on the evaluator's modeling assumptions, underscoring the need for transparent, theory-driven evaluation practices. We conclude by offering an Evaluation Card to help researchers document, justify, and scrutinize the modeling decisions underlying AI evaluations.
arXiv:2509.19824v2 Announce Type: replace
Abstract: Tube-based Model Predictive Control (MPC) is a widely adopted robust control framework for constrained linear systems under additive disturbance. The paper is focused on reducing the numerical complexity associated with the tube parameterization, described as a sequence of elastically-scaled zonotopic sets. A new class of scaled-zonotope inclusion conditions is proposed, alleviating the need for a priori specification of certain set-containment constraints and achieving significant reductions in complexity. A comprehensive complexity analysis is provided for both the polyhedral and the zonotopic setting, illustrating the trade-off between an enlarged domain of attraction and the required computational effort. The proposed approach is validated through extensive numerical experiments.
arXiv:2509.20881v2 Announce Type: replace
Abstract: Code retrieval aims to find relevant code snippets matching natural language queries within massive codebases, playing a vital role in software development. Recent advances leverage PLMs to bridge the semantic gap between natural language (NL) and programming languages (PL), significantly outperforming traditional information retrieval and early deep learning approaches. However, existing methods still face key challenges, including a fundamental semantic gap between human intent and machine execution logic, and limited robustness to diverse code styles. To address this, we propose PseudoBridge, a novel code retrieval framework that introduces pseudo-code as an intermediate, semi-structured modality to align NL semantics with PL logic. Specifically, PseudoBridge consists of two stages: First, we employ an LLM to synthesize pseudo-code, enabling explicit alignment between NL queries and pseudo-code. Second, we introduce a logic-invariant code style augmentation strategy, employing the LLM to generate stylistically diverse yet logically equivalent code implementations, and then align these varied code styles with pseudo-code to enhance robustness. We evaluate PseudoBridge across 10 PLMs and 6 mainstream programming languages. Extensive experiments demonstrate that PseudoBridge consistently outperforms baselines, achieving significant improvements in generalization, particularly in zero-shot scenarios like Solidity and XLCoST. Extended evaluations using open-source LLMs and advanced embeddings confirm that these gains stem from PseudoBridge's intrinsic design, independent of specific closed-source models. PseudoBridge achieves performance comparable to SOTA embedding methods, highlighting the effectiveness of explicit logical and semantic alignment via pseudo-code as a robust solution for code retrieval.
arXiv:2509.21092v2 Announce Type: replace
Abstract: This study investigates how the majority group influences individual judgment formation and expression in anonymous, spontaneous online conversations. Drawing on theories of social conformity and anti-conformity, we analyze everyday dilemmas discussed on social media. First, using digital traces to operationalize judgments, we measure the conversations' disagreement and apply Bayesian regression to capture shifts of judgments formation before and after the group's exposure. Then we analyze changes in judgment expression with a linguistic analysis of the motivations associated with each judgment. Results show anti-conformity behaviors: individuals preserve the majority's positive or negative orientation of judgments but diverge from its stance, with persuasive language increasing post-disclosure. Our findings highlight how online environments reshape social influence compared to offline contexts.
arXiv:2509.21820v2 Announce Type: replace
Abstract: In this paper, we introduce a combination of novel and exciting tasks: the solution and generation of linguistic puzzles. We focus on puzzles used in Linguistic Olympiads for high school students. We first extend the existing benchmark for the task of solving linguistic puzzles. We explore the use of Large Language Models (LLMs), including recent state-of-the-art models such as OpenAI's o1, for solving linguistic puzzles, analyzing their performance across various linguistic topics. We demonstrate that LLMs outperform humans on most puzzles types, except for those centered on writing systems, and for the understudied languages. We use the insights from puzzle-solving experiments to direct the novel task of puzzle generation. We believe that automating puzzle generation, even for relatively simple puzzles, holds promise for expanding interest in linguistics and introducing the field to a broader audience. This finding highlights the importance of linguistic puzzle generation as a research task: such puzzles can not only promote linguistics but also support the dissemination of knowledge about rare and understudied languages.
arXiv:2509.22510v3 Announce Type: replace
Abstract: Alignment of Large Language Models (LLMs) is the ability to satisfy desired objectives during generation, which is critical for trustworthy deployment. In practice, alignment is often operationalized through multiple objectives such as Helpfulness, Harmlessness, and Honesty (HHH). Prior works study alignment via steering vectors in standard Transformer decoders but treat objectives in isolation, where optimizing a single objective can overwrite others, leading to interference. Recent works attempt to address this limitation by extending steering to a 1-to-N Transformer setting by replicating representations into objective-specific pathways, but apply transformations independently, resulting in inconsistent responses across objectives. Similarly, approaches such as safe RLHF and MoE-based designs study trade-offs across objectives but do not constrain objective-specific transformations within a shared representation during inference. As a result, even aligned State-of-the-Art (SOTA) LLMs can struggle to jointly satisfy HHH objectives in complex settings. To address this, we propose Adaptive Multi-Branch Steering (AMBS), a two-stage framework in a 1-to-N Transformer setting that parameterizes objective-specific transformations relative to a shared representation. In Stage I, a shared hidden representation is computed once. In Stage II, this representation is replicated into N pathways and updated relative to a shared reference, capturing objective-specific deviations while restricting divergence. This produces N objective-specific responses within a single forward pass, which can be combined at decoding to obtain a single response across objectives. Across multiple backbones, AMBS improves performance across HHH, with consistent gains in WR, TI, and SS (e.g., Avg 56.5% on LLaMA-2-7B) while maintaining efficiency (e.g., 189 Tok/s, 9 GPU-hrs).
arXiv:2511.19001v2 Announce Type: replace
Abstract: This work presents an innovative computational study of domain-based charge transfer that leverages the localized orbitals of pair coupled cluster doubles (pCCD). This method enables both directional monitoring and quantitative assessment of charge transfer among donor (D), bridge (B), and acceptor (A) moieties. We applied this approach to a series of newly designed carbazole-based prototypical organic dyes, doping the bridge at positions 1, 2, and 3 with nitrogen, oxygen, and sulfur atoms to generate mono-, di-, and tri-doped variants. Our results demonstrate a clear and progressive enhancement in charge transfer as the degree of nitrogen or oxygen doping increases from mono- to di- to tri-doped systems. For mono-doped dyes, the highest forward charge transfer from donor to bridge to acceptor (D$\xrightarrow{}$B$\xrightarrow{}$A) occurs when a heteroatom (N or O) is placed in the terminal ring of the bridge, closer to the acceptor. In di-doped dyes, the largest forward charge transfer is observed when heteroatoms occupy both terminal positions, with one atom (N or S) adjacent to the donor and the other (N) near the acceptor. Nitrogen-doped systems consistently outperform their oxygen and sulfur counterparts. Among all variants, the organic dye doped with three nitrogen atoms at the bridge exhibits the most efficient and highest directional donor-to-acceptor charge transfer (42.6%), making it the most promising candidate for potential applications in dye-sensitized solar cells. Finally, our calculations predict weak charge separation in all systems, indicating that charge transfer predominantly occurs from the bridge to the acceptor.
arXiv:2605.17938v1 Announce Type: new
Abstract: Training data attribution (TDA) should enable generative model interpretability and foster a variety of related downstream tasks. Nonetheless, current TDA approaches lack reliability and robustness, preventing their adoption in real-world setups. In this paper, we take a decisive step towards more reliable and robust TDA for diffusion models. We propose to perform TDA with mirrored unlearning and noise-consistent skew (MUCS). The idea is to fine-tune a second model with bounded mirrored gradient ascent, and to measure the normalized skew of this model with respect to the original one using consistent noise samples. We show that, while being conceptually simple and generic, MUCS systematically outperforms existing methods on three different datasets by a large margin. We additionally study the effect that core design choices have on final performance, and analyze novel aspects regarding the overlap of influential instances across generated items and the potential of ensembling TDA approaches. We believe that our findings may have broader implications for more general unlearning setups, as well as for tasks requiring the comparison of diffusion losses.
arXiv:2511.19078v2 Announce Type: replace
Abstract: Large language models (LLMs) have demonstrated impressive capabilities in natural language understanding and generation, including multi-step reasoning such as mathematical proving. However, existing approaches often lack an explicit and dynamic mechanism to structurally represent and evolve intermediate reasoning states, which limits their ability to perform context-aware theorem selection and iterative conclusion generation. To address these challenges, we propose GraphMind, a novel dynamic graph-based framework that integrates the graph neural network (GNN) with LLMs to iteratively select theorems and generate intermediate conclusions for multi-step reasoning. Our method models the reasoning process as a heterogeneous evolving graph, where nodes represent conditions, theorems, and conclusions, while edges capture logical dependencies between nodes. By encoding the current reasoning state with GNN and leveraging semantic matching for theorem selection, our framework enables context-aware, interpretable, and structured reasoning in a closed-loop manner. Experiments on various question-answering (QA) datasets demonstrate that our proposed GraphMind method achieves consistent performance improvements and significantly outperforms existing baselines in multi-step reasoning, validating the effectiveness and generalizability of our approach.
arXiv:2511.19320v2 Announce Type: replace
Abstract: Preserving first-frame identity while ensuring precise motion control is a fundamental challenge in human image animation. The Image-to-Motion Binding process of the dominant Reference-to-Video (R2V) paradigm overlooks critical spatio-temporal misalignments common in real-world applications, leading to failures such as identity drift and visual artifacts. We introduce SteadyDancer, an Image-to-Video (I2V) paradigm-based framework that achieves harmonized and coherent animation and is the first to ensure first-frame preservation robustly. Firstly, we propose a Condition-Reconciliation Mechanism to harmonize the two conflicting conditions, enabling precise control without sacrificing fidelity. Secondly, we design Synergistic Pose Modulation Modules to generate an adaptive and coherent pose representation that is highly compatible with the reference image. Finally, we employ a Staged Decoupled-Objective Training Pipeline that hierarchically optimizes the model for motion fidelity, visual quality, and temporal coherence. Experiments demonstrate that SteadyDancer achieves state-of-the-art performance in both appearance fidelity and motion control, while requiring significantly fewer training resources than comparable methods. The model has been publicly released at \url{https://mcg-nju.github.io/steadydancer-web}.
arXiv:2511.19953v2 Announce Type: replace
Abstract: Accurate nuclear instance segmentation is a pivotal task in computational pathology, supporting data-driven clinical insights and facilitating downstream translational applications. While large vision foundation models have shown promise for zero-shot biomedical segmentation, most existing approaches still depend on dense supervision and computationally expensive fine-tuning. Consequently, training-free methods present a compelling research direction, yet remain largely unexplored. In this work, we introduce SPROUT, a fully training- and annotation-free prompting framework for nuclear instance segmentation. SPROUT leverages histology-informed priors to construct slide-specific reference prototypes that mitigate domain gaps. These prototypes progressively guide feature alignment through a partial optimal transport scheme. The resulting foreground and background features are transformed into positive and negative point prompts, enabling the Segment Anything Model (SAM) to produce precise nuclear delineations without any parameter updates. Extensive experiments across multiple histopathology benchmarks demonstrate that SPROUT achieves competitive performance without supervision or retraining, establishing a novel paradigm for scalable, training-free nuclear instance segmentation in pathology.
arXiv:2511.20353v3 Announce Type: replace
Abstract: Reasons for mapping an unknown environment with autonomous robots are wide-ranging, but in practice, they are often overlooked when developing planning strategies. Rapid information gathering and comprehensive structural assessment of buildings have different requirements and therefore necessitate distinct methodologies. In this paper, we propose a novel modular Next-Best-View (NBV) planning framework for aerial robots that explicitly uses a reconstruction quality objective to guide the exploration planning. In particular, our approach introduces new and efficient methods for view generation and selection of viewpoint candidates that are adaptive to the user-defined quality requirements, fully exploiting the uncertainty encoded in a Truncated Signed Distance field (TSDF) representation of the environment. This results in informed and efficient exploration decisions tailored towards the predetermined objective. Finally, we validate our method via extensive simulations in realistic environments. We demonstrate that it successfully adjusts its behavior to the user goal while consistently outperforming conventional NBV strategies in terms of coverage, quality of the final 3D map and path efficiency.
arXiv:2509.13045v4 Announce Type: replace-cross
Abstract: Boron-doped diamond crystals (BDD, C$_{1-x}$B$_{x}$) exhibit exceptional mechanical strength, electronic tunability, and resistance to radiation damage. This makes them promising materials for use in gamma-ray crystal-based light sources. To better understand and quantify the structural distortions introduced by doping, which are critical for maintaining channelling efficiency, we perform atomistic-level molecular dynamics simulations on periodic C$_{1-x}$B$_{x}$ systems of various sizes. These simulations allow the influence of boron concentration on the lattice constant and the (110) and (100) inter-planar distances to be evaluated over the concentration range from pure diamond (0%) to 5% boron at room temperature (300 K). Linear relationships between both lattice constant and inter-planar distance with increasing dopant concentration are observed, with a deviation from Vegard's Law. This deviation is larger than that reported by other theoretical and computational studies; however, this may be attributed to an enhanced crystal quality over these studies, a vital aspect when considering gamma-ray crystal light source design. The methodology presented here incorporates several refinements to closely reflect the conditions of microwave plasma chemical vapour deposition (MPCVD) crystal growth. Validation of the methodology is provided through a comprehensive statistical analysis of the structure of our generated crystals. These results enable reliable atomistic modelling of doped diamond crystals and support their use in the design and fabrication of periodically bent structures for next-generation gamma-ray light source technologies.
arXiv:2605.17877v1 Announce Type: new
Abstract: A significant hurdle for current LLMs is the execution of complex, multi-stage tasks. Group Relative Policy Optimization (GRPO) has been emerging as a leading choice, but its reliance on sparse outcome rewards severely limits credit assignment across intermediate steps. Existing remedies such as running full rollouts to assign step-level advantages, calling external LLM judges at each step, or computing intrinsic rewards that require ground-truth answers at every evaluation introduce significant costs or practical constraints. We hypothesize that internal correctness probing over LLM hidden states can be repurposed as a step-level reward signal, potentially addressing all of these limitations at once. However, existing probing research assumes clean inputs, and we first show that this assumption breaks down in multi-step settings: hidden-state probes degrade severely under prefix contamination tracking coherence with the (possibly corrupted) prefix rather than grounded correctness, while attention-based features remain robust to contamination but underperform on clean prefixes. Building on this complementary relationship, we propose the Prefix-Aware Internal Reward (PAIR), a two-stage model with a frozen hidden-state probe estimating belief-consistency and a lightweight attention-based head correcting it toward grounded correctness. Experimental results show that PAIR achieves the highest AUROC on contaminated trajectories while operating at negligible inference cost, enabling dense step-level reward signals for GRPO training without external model calls, ground-truth dependencies, or full-trajectory rollouts.
arXiv:2509.26597v4 Announce Type: replace
Abstract: Control Barrier Functions (CBFs) provide a powerful framework for ensuring safety in dynamical systems. However, their application typically relies on full state information, which is often violated in real-world due to the availability of partial state information. In this work, we propose a neural network-based framework for the co-design of a safety controller, observer, and CBF for partially observed continuous-time systems with input constraints. By formulating barrier conditions over an augmented state space, our approach ensures safety without requiring bounded estimation errors or handcrafted barrier functions. All components are jointly trained by formulating appropriate loss functions, and we introduce a validity condition to provide formal safety guarantees beyond the training data. Finally, we demonstrate the effectiveness of the proposed approach through several case studies.
arXiv:2605.17772v1 Announce Type: new
Abstract: Physical adversarial attacks often overfit single surrogate models and optimization objectives. While ensemble attacks can mitigate this, existing methods struggle with severe gradient conflicts within restricted physical texture spaces, significantly degrading cross-model transferability. To bridge this gap, this paper proposes a Joint Multi-Objective and Multi-Model Optimization Framework (JMOF) that leverages quantitative similarity analysis to select the optimal surrogate model ensemble. Within JMOF, a dual-level mechanism jointly suppresses prediction outputs and flattens intermediate feature distributions, balancing attack efficiency with deep generalization. Additionally, an Orthogonal Gradient Alignment (OGA) strategy resolves cross-model gradient conflicts, transforming mutually repulsive gradients into synergistic optimization directions. Extensive simulated and real-world experiments demonstrate that JMOF outperforms state-of-the-art baselines against diverse black-box detectors. Crucially, JMOF exhibits substantial cross-vision-task generalization, generating attacks capable of simultaneously deceiving object detection and semantic segmentation or monocular depth estimation models. This research advances the generalization limits of physical adversarial attacks, providing a robust framework for evaluating visual AI vulnerabilities in real-world deployments.
arXiv:2605.17928v1 Announce Type: new
Abstract: Transfer Learning, a technique where a model/agent can use the knowledge/expertise that it gained from one task and exploit that to solve another closely-related task, is often used in tackling problems in deep learning. Through this project, we explore transfer learning in the purview of deep reinforcement learning. Specifically, we want to use transfer learning to achieve the fast lap times in OpenAI's Car racing environment by training the agent on one circuit, and racing it on other customized target environments by zero-shot transfer or by additional fine-tuning. In addition, we compare the performance of model-based and model-free approaches, and observe that model-based approaches dominate in performance and converge faster than model-free approaches in this environment. We observe that transfer learning in most setups not only boosts the performance on the target domain, but also shows high performance ability during learning.
arXiv:2605.17879v1 Announce Type: new
Abstract: Training frontier-scale foundation models involves coordinating tens of thousands of GPUs over multi-month runs, where even minor performance degradations can accumulate into substantial efficiency losses. Existing health-check mechanisms, such as NCCL tests or GPU burn-in, primarily focus on functional correctness and often fail to detect fail-slow behaviors that silently degrade system performance. In this paper, we present Guard, a scalable system for detecting stragglers and ensuring node health in large-scale training clusters. Guard combines lightweight online performance monitoring during training with an offline node-sweep mechanism that systematically evaluates and qualifies nodes before they participate in production workloads. This design enables Guard to detect both acute failures and long-running fail-slow behaviors that traditional diagnostics cannot capture. Deployed on large-scale foundation model pretraining workloads, Guard improves mean FLOPs utilization by up to 1.7x, reduces run-to-run training step variance from 20% to 1%, increases mean time to failure (MTTF), and significantly reduces operational and debugging overhead. These results demonstrate that proactive straggler detection and systematic node qualification are critical for maintaining stable and efficient large-scale training.
arXiv:2605.17787v1 Announce Type: new
Abstract: It is widely believed that stochastic gradient descent (SGD) performs significantly worse than adaptive optimizers such as Adam in pre-training Large Language Models (LLMs). Yet the underlying reason for this gap remains unclear. In this work, we attribute a large part of the discrepancy to SGD's inability to sustain learning rates comparable to Adam's much larger effective learning rates. Through empirical and theoretical analysis of LLM pre-training dynamics, we identify that training is characterized by small gradient norms and large weight-to-gradient ratios, an effect that becomes more pronounced with larger batch sizes typical in pre-training, necessitating such large effective learning rates. However, we find that output-layer gradient magnitudes become highly uneven across token classes, and that large gradient spikes frequently occur during training. Together, these effects severely restrict the admissible learning rate of SGD. Guided by this understanding, we show that simple clipping mechanisms that stabilize SGD at large learning rates enable it to recover most of Adam's performance. In our large-scale experiments, the validation loss gap between large-learning-rate SGD and Adam shrinks from more than 50% to only about 3.5% when pre-training a 1B-parameter LLaMA model with a 1M-token batch size.
arXiv:2605.17791v1 Announce Type: new
Abstract: Wireless resource allocation in digital-twin-enabled unmanned aerial vehicle (UAV) swarms must be both network-feasible and certifiably safe for closed-loop control.
Existing packet-level or scalar-priority schedulers cannot meaningfully compare heterogeneous multi-hop actions that differ simultaneously in route, retransmission depth, blocklength, bidirectional delay, delivery probability, and TDMA slot cost.
This paper introduces a certificate-guided resource allocation framework for low-altitude multi-hop UAV swarms.
A digital twin maps predicted topology, channel, route, and controller-side state into a shared five-dimensional quality-of-service (QoS) certificate comprising uplink/downlink delay bounds, directional delivery guarantees, and a certified upper bound on the interval between successful bidirectional interactions.
A state-conditioned stochastic drift test then admits only certificates whose augmented Lyapunov drift is nonpositive under the current controller state.
Admitted actions are reduced to certified supply frontiers by removing dominated route-slot configurations, and the online scheduler maximizes Lyapunov-drift reduction under a shared TDMA slot budget via exact dynamic programming.
Closed-loop ns-3 simulations demonstrate that the proposed framework outperforms fixed-service, certificate-filtered fixed-priority, dynamic-transmission-count, and value-of-information baselines in both tracking accuracy and high-risk state suppression under identical communication budgets.
arXiv:2605.18410v1 Announce Type: new
Abstract: Identifying which newly published scientific papers are likely to become highly cited is important for prioritizing research attention, supporting editorial decisions, and guiding the allocation of scientific resources, particularly under cold-start conditions where little direct evidence is available at publication time. In this work, we formulate impact prediction as a cohort-normalized top-P% classification task and compare graph-based and LLM-based approaches under a unified framework. We construct citation and textual-similarity graphs under temporal constraints and generate Node2Vec representations, either alone or combined with OpenAI text embeddings. The best supervised configuration combines directed citation graphs with textual embeddings, reaching approximately 0.84-0.85 AUC. We also evaluate a GPT-based GraphRAG setup, using GPT 5.5 and 5.4 Nano, in which graph neighborhoods are used as contextual evidence for prediction. Although the LLM-based approach achieves high performance, retrieved context does not consistently improve results; target-only prompts often perform as well as or better than GraphRAG prompts achieving the 0.87 mark. These findings indicate that structural and textual signals are complementary for supervised prediction, while retrieval augmentation must be carefully evaluated against simpler LLM baselines.
arXiv:2605.18572v1 Announce Type: new
Abstract: Persuasive dialogue generation plays a vital role in decision-making, negotiation, counseling, and behavior change, yet it remains a challenging problem. In complex persuasion where the persuadee's internal states are not expressed clearly, the persuader must interpret responses, infer the persuadee's latent mental states (e.g., beliefs and desires), and translate them into targeted, strategy-consistent actions; however, current approaches often produce generic or weakly grounded responses even when such cues are identified. Moreover, although large language models (LLMs) can generate persuasive content, their performance varies substantially across domains due to uneven knowledge coverage and limited reasoning generalization. To address these challenges, we propose MA$^{2}$P, a meta-cognitive autonomous intelligent agent framework for complex persuasion. Specifically, we develop an autonomous multi-agent architecture that coordinates perception management, mental-state inference, strategy execution, memory maintenance, and performance evaluation. To mitigate cross-domain performance variation, we further design a meta-cognitive configurator that selects an appropriate meta-strategy from a structured knowledge base at the outset, thereby guiding subsequent reasoning and planning. Experimental results show that our approach achieves a higher persuasion success rate than baselines.
arXiv:2510.02590v2 Announce Type: replace
Abstract: The use of target networks is a popular approach for estimating value functions in deep Reinforcement Learning (RL). While effective, the target network remains a compromise solution that preserves stability at the cost of slowly moving targets, thus delaying learning. Conversely, using the online network as a bootstrapped target is intuitively appealing, albeit well-known to lead to unstable learning. In this work, we aim to obtain the best out of both worlds by introducing a novel update rule that computes the target using the MINimum estimate between the Target and Online network, giving rise to our method, MINTO. Through this simple, yet effective modification, we show that MINTO enables faster and stable value function learning, by mitigating the potential overestimation bias of using the online network for bootstrapping. Notably, MINTO can be seamlessly integrated into a wide range of value-based and actor-critic algorithms with a negligible cost. We evaluate MINTO extensively across diverse benchmarks, spanning online and offline RL, as well as discrete and continuous action spaces. Across all benchmarks, MINTO consistently improves performance, demonstrating its broad applicability and effectiveness.
arXiv:2510.03879v3 Announce Type: replace
Abstract: Translating C to memory-safe languages, like Rust, prevents critical memory safety vulnerabilities that are prevalent in legacy C software. Even with recent LLM-based and tool-augmented translators, the resulting Rust code frequently diverges from the C source on inputs absent from the test suite used during translation; this correctness gap on unseen inputs remains a dominant obstacle to reliable, automatic C-to-Rust translation. In this work, we present ACToR (Adversarial C To Rust), a simple LLM-agent loop that closes this gap by adversarially searching for inputs on which the translation diverges from the C source, and using them to drive subsequent refinements. Inspired by GANs, ACToR pits a translator agent against a discriminator agent that collaborate to iteratively refine the Rust translation. On each iteration, the translator agent synthesizes and refines a Rust translation to pass an existing suite of tests, and then the discriminator agent finds new failing tests by constructing and refining a differential fuzzer over the C and Rust binaries. Across 63 real-world command-line C utilities, with an average size of 473 lines of code and the longest reaching thousands of lines in size, ACToR achieves over 90% test pass rate with zero human intervention. The improvement holds across seven agent-LLM configurations on our micro-benchmark, indicating that the loop is largely independent of the choice of underlying translator and LLM. Compared to a non-adversarial, coverage-driven test-generation baseline, ACToR improves correctness by up to 36.7%. When applied on top of one recent translator, C2SaferRust, ACToR further improves the validation pass rate by 16.6%.