arXiv:2606.04555v1 Announce Type: new
Abstract: Long-horizon conversational agents need to interact with users through evolving events, tasks, and goals. Such histories are naturally temporal, yet many existing memory systems organize information primarily by topical similarity and may ignore the order in which events occur. We introduce Segment Tree Memory, or SegTreeMem, a memory architecture that represents conversation history as a temporally ordered Segment Tree over utterances. SegTreeMem incrementally inserts new utterances through an online rightmost-frontier update rule, preserving chronological order while forming hierarchical memory segments. For retrieval, SegTreeMem propagates relevance scores through the tree to combine local semantic matching with hierarchical temporal context. Across three long-horizon memory benchmarks and two LLM backbones, SegTreeMem improves answer quality over flat retrieval, graph-structured memory, and tree-structured memory baselines. Additional temporal-order permutation analysis shows that the performance gain depends on preserving temporal order during memory construction, supporting the claim that temporal order is a key structure for agentic memory.
Science Journals
arXiv:2502.03799v4 Announce Type: replace
Abstract: Large Language Models (LLMs) are prone to generating plausible yet incorrect responses, known as hallucinations. Effectively detecting hallucinations is therefore crucial for the safe deployment of LLMs. Recent research has linked hallucinations to model uncertainty, suggesting that hallucinations can be detected by measuring dispersion over answer distributions obtained from multiple samples drawn from a model. While drawing from the distribution over tokens defined by the model is a natural way to obtain samples, in this work, we argue that it is suboptimal for the purpose of detecting hallucinations. We show that detection can be improved significantly by taking into account model uncertainty in the Bayesian sense. To this end, we propose a very simple, training-free approach based on perturbing an appropriate subset of model parameters, or equivalently hidden unit activations, during sampling. We demonstrate that our approach significantly improves inference-time hallucination detection over standard sampling across diverse datasets, model architectures, and uncertainty metrics.
arXiv:2502.02718v2 Announce Type: replace
Abstract: The paper studies parametric Reduced Order Models (ROMs) for the Kuramoto--Sivashinsky (KS) and generalized Kuramoto--Sivashinsky (gKS) equations. We consider several POD and POD-DEIM projection ROMs with various strategies for parameter sampling and snapshot collection. The aim is to identify an approach for constructing a ROM that is efficient across a range of parameters, encompassing several regimes exhibited by the KS and gKS solutions: weakly chaotic, transitional, and quasi-periodic dynamics. We describe such an approach and demonstrate that it is essential to develop ROMs that adequately represent the short-time transient behavior of the gKS model.
arXiv:2601.19449v2 Announce Type: replace
Abstract: Graph neural networks (GNNs) are widely believed to excel at node representation learning through trainable neighborhood aggregations. We challenge this view by introducing Fixed Aggregation Features (FAFs), a training-free approach that transforms graph learning tasks into tabular problems. This simple shift enables the use of well-established tabular methods, offering strong interpretability and the flexibility to deploy diverse classifiers. Across 14 benchmarks, well-tuned multilayer perceptrons trained on FAFs rival or outperform state-of-the-art GNNs and graph transformers on 12 tasks -- often using only mean aggregation. The only exceptions are the Roman Empire and Minesweeper datasets, which typically require unusually deep GNNs. To explain the theoretical possibility of non-trainable aggregations, we connect our findings to Kolmogorov-Arnold representations and discuss when mean aggregation can be sufficient. In conclusion, our results call for (i) richer benchmarks benefiting from learning diverse neighborhood aggregations, (ii) strong tabular baselines as standard, and (iii) employing and advancing tabular models for graph data to gain new insights into related tasks.
arXiv:2502.01711v3 Announce Type: replace
Abstract: Symmetry is an important inductive bias that can improve model robustness and generalization across many deep learning domains. In multi-agent settings, a priori known symmetries have been shown to address a fundamental coordination failure mode known as mutually incompatible symmetry breaking; e.g. in a game where two independent agents can choose to move "left'' or "right'', and where a reward of +1 or -1 is received when the agents choose the same action or different actions, respectively. However, the efficient and automatic discovery of environment symmetries, in particular for decentralized partially observable Markov decision processes, remains an open problem. Furthermore, environmental symmetry breaking constitutes only one type of coordination failure, which motivates the search for a more accessible and broader symmetry class. In this paper, we introduce such a broader group of previously unexplored symmetries, which we call expected return symmetries, which contains environment symmetries as a subgroup. We show that agents trained to be compatible under the group of expected return symmetries achieve better zero-shot coordination results than those using environment symmetries. As an additional benefit, our method makes minimal a priori assumptions about the structure of their environment and does not require access to ground truth symmetries.
arXiv:2606.04557v1 Announce Type: new
Abstract: Large Language Models can reason over long contexts, yet prefilling millions of tokens is wasteful as much of the content remains static across queries. Cartridges address this by distilling document collections into reusable key-value (KV) caches that eliminate prefilling while preserving accuracy. A critical limitation of this approach is that cartridges are monolithic and non-compositional: encoding an entire collection into a single KV block does not scale, and naively mixing cartridges trained in isolation collapses performance to near chance. We introduce Cartridges at Scale (CAS), a training framework for scalable multi-cartridge learning with dynamic distractor mixing and a memory-efficient budget manager that rotates hundreds of per-document cartridges between GPU and persistent storage. Our approach scales to collections exceeding a million tokens, improving over a monolithic cartridge by 10-31 points at comparable token budgets. Oracle cartridge accuracy falls within 2-6 points of full in-context learning even at high compression. When paired with retrieval for cartridge selection, CAS matches or exceeds conventional RAG accuracy while consuming 3-4x fewer prompt tokens.
arXiv:2606.04563v1 Announce Type: new
Abstract: Computing is accompanied by both positive and negative commons throughout its lifecycle of creation, execution, and disposal. We examine two governance systems situated within this lifecycle -- global e-waste trade and the Linux kernel community -- to evaluate whether Elinor Ostrom's eight design principles for common-pool resource (CPR) governance extend to the management of negative common-pool resources (NCPRs). Unlike traditional CPRs where communities work to preserve a finite resource (i.e. clean water), NCPR governance seeks to collectively reduce a negative shared stock. In our two cases, e-waste governance aims to reduce the volume of mismanaged waste and illicit trade, while the Linux community aims to reduce the number of error-prone or malicious contributions that reach the main branch and, in turn, extend the life of existing hardware. Through qualitative analysis of primary sources from each domain, we find that the same eight principles by Ostrom that aid positive commons governance tend to appear in successful negative commons governance systems. We argue that future NCPR governance design should prioritize Ostrom's principles, particularly clearly defined boundaries and well-functioning nested structures.
arXiv:2502.01576v2 Announce Type: replace
Abstract: Multi-modal Large Language Models (MLLMs) excel in vision-language tasks but remain vulnerable to visual adversarial perturbations that can induce hallucinations, manipulate responses, or bypass safety mechanisms. Existing methods seek to mitigate these risks by applying constrained adversarial fine-tuning to CLIP vision encoders on ImageNet-scale data, ensuring their generalization ability is preserved. However, this limited adversarial training restricts robustness and broader generalization. In this work, we explore an alternative approach of leveraging existing vision classification models that have been adversarially pre-trained on large-scale data. Our analysis reveals two principal contributions: (1) the extensive scale and diversity of adversarial pre-training enables these models to demonstrate superior robustness against diverse adversarial threats, ranging from imperceptible perturbations to advanced jailbreaking attempts, without requiring additional adversarial training, and (2) end-to-end MLLM integration with these robust models facilitates enhanced adaptation of language components to robust visual features, outperforming existing plug-and-play methodologies on complex reasoning tasks. Through systematic evaluation across visual question-answering, image captioning, and jail-break attacks, we demonstrate that MLLMs trained with these robust models achieve superior adversarial robustness while maintaining favorable clean performance. Our framework achieves 2x and 1.5x average robustness gains in captioning and VQA tasks, respectively, and delivers over 10% improvement against jailbreak attacks. Code and pretrained models will be available at https://github.com/HashmatShadab/Robust-LLaVA.
arXiv:2603.21107v2 Announce Type: replace
Abstract: The performance of low-power wireless sensing networks can be influenced by both external environmental factors and internal imperfections which often arise due to manufacturing tolerance during mass production. Understanding the conditions and extent of these influences is important not only to achieve high performance and high energy efficiency, but also to carry our environment and radio specific configurations. In this paper we demonstrate, through extensive practical deployments and experiments, the extent to which external and internal factors affect the link quality of low-power wireless sensor networks. Moreover, we propose a lightweight statistical outlier detection technique and define all the parameter thereof in terms of the statistics of both the raw and the predicted link quality metrics (RSSI). Our study considers more than 15 different physical environments consisting of rivers, lakes, bridges, forests, and gardens, as well as four widely employed heterogeneous low-power radios.
arXiv:2605.12226v4 Announce Type: replace
Abstract: Recent advances in large language models (LLMs) pose new challenges for ontology matching (OM). While OM systems built on LLMs have shown remarkable capabilities in discovering more matching candidates, traditional OM validation that relies on domain experts has become overwhelming. In this study, we explore the use of crowdsourcing for OM validation and introduce a novel crowdsourcing system. We propose three domain-specific mechanisms, namely differential trustworthiness, coherence pre-filling, and time-dependent opinion, to ensure the quality of crowdsourcing for OM validation. We demonstrate that our crowdsourcing system can be integrated with existing OM systems to enable human-in-the-loop validation. The evaluation of the system shows its effectiveness in handling diverse user groups and different annotation settings. We discuss two real-world use cases of the system and current limitations for improvement.
arXiv:2501.16962v2 Announce Type: replace
Abstract: Modern computing systems rely on the Unified Extensible Firmware Interface (UEFI), which has replaced the legacy Basic Input/Output System (BIOS) as the firmware standard for the modern boot process. Although the UEFI represents a significant advancement in system firmware, it is increasingly targeted by threat actors seeking to exploit its execution environment and take advantage of its persistence mechanisms. While some security-related analysis of UEFI components has been performed--primarily via debugging and runtime behavior testing--to the best of our knowledge, no prior study has specifically addressed the capturing and analysis of volatile UEFI runtime memory to detect malicious exploitation during the pre-OS phase. This gap in UEFI forensic tools limits the ability to conduct in-depth security analysis in pre-OS environments. Such a gap is particularly surprising, given that memory forensics is widely regarded as foundational to modern incident response, as reflected by the popularity of above-OS memory analysis frameworks, such as Rekall, Volatility, and MemProcFS. To address the lack of below-OS memory forensics, we introduce a framework for UEFI memory forensics. The proposed framework consists of two components: UEFIMemDump, a memory acquisition tool, and UEFIDumpAnalysis, an extendable collection of analysis modules capable of detecting malicious activities such as function pointer hooking, inline hooking, malicious image loading, and gadget-based control-flow manipulation. Our proof-of-concept implementation demonstrates the framework's ability to detect modern UEFI threats, such as Thunderstrike, CosmicStrand, and Glupteba bootkits. By providing an open-source solution, our work enables researchers and practitioners to investigate firmware-level threats, develop additional analysis modules, and advance overall below-OS security through UEFI memory analysis.
arXiv:2606.04017v1 Announce Type: new
Abstract: Long-running AI agents fail not only when inference fails or tools are underspecified, but when independently evolving model and harness layers change the semantics of belief, capability, and goal commitments across their boundary - a failure class this paper terms Interface Volatility. This paper argues that Agent Epistemic Integrity (AEI) must be treated as a first-class architectural constraint, achievable only through joint model-harness design organized around an explicit interface contract. The central claim is that the model-harness interface contract is the precondition for joint design; its operational form is a four-level hierarchy - goal validity, action-archetype sequencing, tool-instance selection, and invocation-level failure discrimination - that specifies what the boundary must preserve and what structured outputs the model must return for the contract to hold across levels. This reframes long-running agent design away from flat action loops and toward contract-preserving control over persistent state. Evaluation and training should therefore derive from the contract itself, testing whether belief, tool, and goal commitments hold across session boundaries and independent layer upgrades.
arXiv:2412.20803v2 Announce Type: replace
Abstract: Event cameras are biologically inspired sensors garnering significant attention from both industry and academia. Mainstream methods favor frame and voxel representations, which reach a satisfactory performance while introducing time-consuming transformations, bulky models, and sacrificing fine-grained temporal information. Alternatively, Point Cloud representation demonstrates promise in addressing the mentioned weaknesses, but it has limited scalability in abstracting features of higher spatial resolution and longer temporal sequence events. In this paper, we propose a Scalable Network named SECNet to leverage Event Cloud representation. SECNet integrates polarity at the structural level by innovating the Event-based Group and Sampling module rather than only at the input level. To accommodate the surge in the number of events, SECNet embraces feature extraction in the frequency domain via the Fourier transform.This approach not only substantially extinguishes the explosion of Multiply Accumulate Operations but also effectively abstracts spatio-temporal features. We conducted extensive experiments on \textbf{ten} event-based datasets, and substantiate the scalability, effectiveness, and efficiency of SECNet. Our code will be available at: https://github.com/rhwxmx/SECNet_ICML.
arXiv:2606.04016v1 Announce Type: new
Abstract: We describe a reproducible computational framework for upper-bound searches on r_3(N), the maximum size of a 3-term-arithmetic-progression-free subset of [1,N]. The framework combines a verified lower-bound witness, endpoint forcing, depth-d witness-variable splitting, OEIS A003002 window-cardinality pruning, and recursive refinement of timed-out subproblems. Applied to the frontier case N = 212, K = 44, it found no feasible 44-set across millions of CP-SAT subproblems, supporting but not proving the conjectural value r_3(212) = 43. A 300-second recap leaves 45 resistant chunks; one-hour HiGHS MIP closes none of them; the full eight-hour HiGHS audit closes 25/45 and leaves 20/45 with dual bounds still pinned at 0.0. A CDCL/SAT re-attack on those LP-paradigm-resistant chunks closes 18 via conflict-driven clause learning; all eighteen carry independently verified DRAT proofs. The remaining two chunks (T1c) resist every tested paradigm under generous wall caps. We release the witness, solver scripts, result logs, tiered benchmark instances, verified DRAT/LRAT proofs, and a Lean formal-proof-search encoding of T1c, and frame the unit-gap problem r_3(212) in {43,44} as a target for stronger additive-combinatorial bounds, custom branch-and-bound, or formal proof-search systems.
arXiv:2606.03938v2 Announce Type: replace
Abstract: Multi-epoch training is becoming the standard now that compute is growing faster than the supply of high-quality text. But pretraining a single model saturates within a few passes, long before the compute budget is exhausted. We argue this calls for a conceptual shift from training a single model toward exploring a population of models and aggregating their predictions. We introduce hyper-epoch pretraining (q0), which turns a multi-epoch budget into a population of diverse models whose combined predictions reach a lower validation loss than a single refined model. q0 reduces to three core primitives. A cyclic schedule with anti-correlated learning rate and weight decay collects diverse models from a few parallel trajectories. Chain distillation trains each model against its predecessor so that model quality compounds across the population. A learned prior, fit on a held out set, selects and weights members for any inference budget. On a 1.8B-parameter model trained on 100M FineWeb tokens, q0 matches a strong 256-epoch ensemble baseline using only ~56 epochs (~4.6x fewer), or ~67 epochs (~3.8x fewer) when matched to the baseline's ensemble size, and continues to improve beyond it. These gains reach cumulative ~12.9x data efficiency under the Slowrun setting and transfer to downstream benchmarks. Crucially, the optimal allocation shifts with the budget, so we give prescriptive recipes for how to spend a given epoch budget to maximize generalization, from a single epoch up to the largest budgets.
arXiv:2606.03937v2 Announce Type: replace
Abstract: While token-level entropy is commonly recognized as effective for credit assignment in text-only reinforcement learning with verifiable rewards (RLVR), it remains unclear whether this mechanism still holds in visual reasoning. Our controlled study shows that this mechanism collapses in visual reasoning due to the omission of vision-sensitive tokens with naturally low entropy. Although existing multimodal RL methods increasingly acknowledge the importance of visual perception, they struggle to satisfy the inherent demand for interleaving precise perceptual grounding with semantic reasoning, either lacking systematic visual measurements or overlooking that token entropy primarily drives semantic exploration. To address this, we introduce VEPO (Vision-Entropy token-selection for Policy Optimization), an effective RL framework explicitly integrating visual sensitivity with token entropy via a principled multiplicative coupling, where VEPO redirects gradient credit toward tokens which are simultaneously visually grounded and highly informative. Extensive experiments demonstrate VEPO's leading performance, significantly outperforming the entropy-only baseline by 2.28 points at 7B-scale and 3.15 points at 3B-scale. Ablations further substantiate the soundness of our method.
arXiv:2511.12784v3 Announce Type: replace
Abstract: Large Language Models (LLMs) have recently emerged as powerful tools for autoformalization. Despite their impressive performance, these models can still struggle to produce grounded and verifiable formalizations. Recent work in text-to-SQL, has revealed that LLMs can be sensitive to paraphrased natural language (NL) inputs, even when high degrees of semantic fidelity are preserved. In this paper, we investigate this claim in the autoformalization domain. Specifically, we evaluate the robustness of LLMs generating formal proofs with semantically similar paraphrased NL statements by measuring semantic and compilation validity. Using the formal benchmarks MiniF2F and Lean 4 version of ProofNet, and two modern LLMs, we generate paraphrased natural language statements and cross-evaluate these statements across both models. The results of this paper reveal performance variability across paraphrased inputs, demonstrating that minor shifts in NL statements can significantly impact model outputs.
arXiv:2603.20884v3 Announce Type: replace
Abstract: To alleviate the heavy burden of paper screening, researchers increasingly rely on existing AI agents, such as AI reviewers or DeepResearch, for paper evaluation and novelty assessment. However, lacking specialized mechanisms for processing scholarly literature, their analyses often produce superficial results with noticeable deficiencies in quality. To bridge this gap, we introduce MemoNoveltyAgent, a multi-agent system designed to generate comprehensive and faithful novelty reports. Beyond retrieving concrete prior-paper evidence via RAG, our system incorporates a high-level abstract memory constructed from large-scale scholarly corpora. This memory organizes research into hierarchical trees to distill field-specific evolutionary trajectories, thereby providing a broader historical context. Furthermore, we decompose papers into discrete novelty points for fine-grained analysis and retrieval, while employing a self-validation mechanism to improve report faithfulness. Finally, to address the evaluation challenges of such open-ended generation tasks, we propose a RAG-augmented checklist evaluation method that enables reliable and evidence-grounded assessments. Extensive experiments demonstrate that MemoNoveltyAgent outperforms GPT-5 DeepResearch by 13.69%. Code and demo are available at https://github.com/SStan1/MemoNoveltyAgent
arXiv:2412.18134v5 Announce Type: replace
Abstract: Randomized self-reductions (RSRs) express $f(x)$ using $f$ evaluated at random correlated points, enabling self-correcting programs, instance-hiding protocols, and applications in complexity theory and cryptography. Yet discovering RSRs has required manual expert derivation for over 40 years, limiting their practical use. We present Bitween for automated RSR learning. First, we formalize RSR learning with sample complexity analysis under correlated sampling. Second, we develop Vanilla Bitween, which integrates multiple backends (linear regression, genetic programming, symbolic regression, and mixed-integer programming). The linear regression backend outperforms the others, discovering RSRs for 43 of 80 functions (54%) in RSR-Bench, our benchmark suite, including the first known reduction for sigmoid. Third, we introduce Agentic Bitween, a neuro-symbolic approach where LLM agents propose novel query functions beyond the fixed set ($x+r$, $x-r$, $x \cdot r$, $x$, $r$) in prior work. Agentic Bitween discovers RSRs for 64 of 80 functions (80%), outperforming pure neural baselines in both RSR discovery and verification accuracy.
arXiv:2606.04072v1 Announce Type: new
Abstract: Deep learning models are increasingly central to autonomous vehicle (AV) pipelines, yet their integration has traditionally followed a monolithic design where perception, planning, and control execute on a single onboard computer. This design overlooks the emerging paradigm of cooperative autonomy, where vehicles interact with roadside units (RSUs), edge servers, and cloud-hosted intelligence through vehicle-to-everything (V2X) connectivity. Cooperative perception and control improve safety and efficiency, but also introduce systems-level challenges: network latency, compute heterogeneity, and multi-tenant contention, all critically affect real-time decision-making. These challenges are further amplified by the increasing reliance on large foundation models, whose scale necessitates cloud deployment.
We present CADET (Cooperative Autonomy through Distributed Experimentation Toolkit), a modular platform for systematic and reproducible evaluation of distributed cooperative autonomy systems under realistic deployment conditions. CADET decouples the AV stack into composable modules that can be flexibly deployed across vehicles, infrastructure, and edge/cloud tiers. The framework integrates state-of-the-art models, incorporates trace-driven network and workload emulation, and provides synchronized model-, system-, and task-level instrumentation. Through V2V and V2I experiments, we show that distributed deployment choices fundamentally shape safety, with V2V intent packets outperforming cloud-based perception and RSU-assisted perception sustaining safety until overloaded by concurrent requests. Although designed for AV pipelines, CADET also supports dataset-driven experimentation, enabling systems and ML researchers to benchmark distributed inference workloads independently of full vehicle simulation. CADET is open source, with code and demo available at https://nesl.github.io/cadet-web.
arXiv:2606.04565v1 Announce Type: new
Abstract: The efficiency of reactive near-field wireless power transfer (WPT) systems degrades rapidly with increasing separation distance and is highly sensitive to misalignment between transmitting and receiving coils. These limitations restrict the mobility of powered devices and confine many near-field WPT applications to static scenarios. To address these challenges, a multiple-input multiple-output (MIMO) WPT configuration is investigated due to its capability to shape the magnetic field distribution between the transmitter and receiver. Maximum power transfer efficiency can be achieved by appropriately setting the amplitude and phase of each transmitting coil; however, determining these optimal settings requires accurate knowledge of the system's S-parameters. This paper presents the use of the Nelder-Mead iterative optimization algorithm to estimate the input amplitude and phase settings that maximize transfer efficiency in a near-field WPT system. The implementation comprises a four-element transmitter and a two-element receiver. Based on measured S-parameters, the proposed approach significantly improves WPT efficiency under both aligned and misaligned conditions.
arXiv:2606.04570v1 Announce Type: new
Abstract: Higher-Order Ambisonics (HOA) encoding from sparse, irregular microphone arrays remains a critical challenge for consumer spatial audio capture in immersive communication and XR. We propose Flow-HOA, a generative framework that jointly optimizes a multi-dimensional objective encompassing time-domain, spectral, and spatial fidelity while producing a deployable, time-invariant bank of Finite Impulse Response (FIR) encoding filters. Using conditional flow matching, the model learns to map a simple prior distribution to the target distribution of FIR filter coefficients. Training is guided by a composite loss that balances time-domain waveform fidelity, multi-resolution spectral consistency, sub-band energy preservation, and spatial directivity constraints. Objective evaluations on synthetically simulated data demonstrate improved performance over strong model-based baselines in both signal fidelity and spatial accuracy metrics. Subjective listening tests on real microphone array recordings further confirm that Flow-HOA yields higher overall sound quality with reduced artifacts, demonstrating generalization from synthetic training data to real-world capture conditions.
Independence and Domination on Bounded-Treewidth Graphs: Integer, Rational, and Irrational Distances
arXiv:2606.04572v1 Announce Type: new
Abstract: The distance-d variants of Independent Set and Dominating Set problems have been extensively studied from different algorithmic viewpoints. In particular, the complexity of these problems are well understood on bounded-treewidth graphs [Katsikarelis, Lampis, and Paschos, Discret. Appl. Math 2022][Borradaile and Le, IPEC 2016]: given a tree decomposition of width t, the two problems can be solved in time $d^t \cdot n^{O(1)}$ and $(2d + 1)t \cdot n^{O(1)}$, respectively. Furthermore, assuming the Strong Exponential-Time Hypothesis (SETH), the base constants are best possible in these running times: they cannot be improved to $d-{\epsilon}$ and $2d+1-{\epsilon}$, respectively, for any ${\epsilon} > 0$. We investigate continuous versions of these problems in a setting introduced by Megiddo and Tamir [SICOMP 1983], where every edge is modeled by a unit-length interval of points. In the {\delta}-Dispersion problem, the task is to find a maximum number of points (possibly inside edges) that are pairwise at distance at least {\delta} from each other. Similarly, in the {\delta}-Covering problem, the task is to find a minimum number of points (possibly inside edges) such that every point of the graph (including those inside edges) is at distance at most {\delta} from the selected point set. We provide a comprehensive understanding of these two problems on bounded-treewidth graphs.
arXiv:2606.04968v1 Announce Type: new
Abstract: Large vision-language-action (VLA) policies are increasingly trained as conditional generative models over action chunks. Yet deployment produces mixed-quality experience-successful demonstrations, partial completions, recoverable mistakes, and failures-that is difficult to use with standard imitation. Full behavior cloning (BC) imitates failures, filtered BC discards useful sub-trajectories, and offline reinforcement learning adds a large critic. We introduce ForesightFlow, a self-guided flow-matching policy that augments each generated action chunk with a learned success-potential trajectory. The same flow proposes and scores candidate actions, enabling best-of-$K$ inference without an external critic. The key issue is that policy improvement and value calibration require different supervision: advantage weighting should emphasize high-quality actions, but applying the same weights to potential coordinates suppresses failure gradients and creates overconfident scores. We address this with decoupled advantage-weighted flow matching, applying exponentiated advantage weights only to action velocities while training potential velocities uniformly. We further derive a one-step boundary estimator for conditional flow matching, allowing advantage computation with a single stop-gradient forward pass. Across five BEHAVIOR-1K simulation tasks and five real-world bimanual tasks, ForesightFlow improves over imitation baselines, matches the strongest separate-critic baseline in simulation success, improves real-world success, and reduces training compute by $38\%$. Ablations show that decoupling prevents value hallucination, the one-step estimator preserves candidate-ranking fidelity, and self-guided sampling improves long-horizon execution.
arXiv:2502.08870v2 Announce Type: replace
Abstract: We provide an approach for the analysis of randomised exploration algorithms like Thompson sampling that does not rely on forced optimism or posterior inflation. With this, we demonstrate that in the $d$-dimensional linear bandit setting, when the action space is smooth and strongly convex, randomised exploration algorithms enjoy an $n$-step regret bound of the order $O(d\sqrt{n} \log(n))$. Notably, this shows for the first time that there exist non-trivial linear bandit settings where Thompson sampling can achieve optimal dimension dependence in the regret.