Academic Intelligence · Curated Daily

探索全球前沿学术脉络

AcademicHub 汇聚顶级期刊与预印本平台的实时文献。定制您的专属科研雷达,利用大语言模型自动生成交叉领域文献分析简报。

01.
Nature (Science) 2026-06-11

Daily briefing: Deep-sea whale graveyard is a treasure trove of fossils

作者:

Researchers have uncovered more than 400 fossilized whale bones in an ocean-floor chasm. Plus, the working lives of scientists, in pictures, and how AI could slow the pace of research publication for the better. Researchers have uncovered more than 400 fossilized whale bones in an ocean-floor chasm. Plus, the working lives of scientists, in pictures, and how AI could slow the pace of research publication for the better.

02.
arXiv (CS.AI) 2026-06-15

StreamMemBench: Streaming Evaluation of Agent Memory for Future-Oriented Assistance

arXiv:2606.14571v1 Announce Type: new Abstract: A central role of personal-agent memory is to turn stored information and prior interactions into future-oriented assistance. In daily use, useful cues come from what the agent observes and how the user interacts with the agent, and the agent must carry them forward from the current request to similar future tasks. Existing memory benchmarks usually test dialogue recall or task improvement in isolation, leaving the trajectory from streaming observations to later assistance largely untested. We introduce StreamMemBench, a streaming benchmark that constructs a two-step task sequence around each evidence anchor from EgoLife egocentric streams. The initial task tests evidence use, while the follow-up task tests whether feedback and interaction experience are reused. Four metrics diagnose evidence recall, initial evidence use, feedback incorporation, and follow-up reuse. Experiments with eight memory systems across two backbones show that current systems often fail to use observed evidence or turn feedback into reliable follow-up behavior, even when evidence is stored or feedback is incorporated locally. StreamMemBench is publicly available at https://github.com/landian60/StreamMemBench.

03.
arXiv (CS.CV) 2026-06-16

A Comprehensive Survey of Medical Image Segmentation: Challenges, Benchmarks, and Beyond

Medical image segmentation plays a critical role in clinical diagnostics, treatment planning, disease monitoring, and neurological disorder identification. This article presents a comprehensive review of its systematic development, covering widely used public datasets, representative methods built on the U-Net, Transformer, and SAM architectures, and key evaluation metrics with their differences, followed by an analysis of major challenges from multiple perspectives. Unlike surveys that focus on a single model family or a specific clinical application, this review organizes U-Net-, Transformer-, and SAM-based methods within a unified analytical framework, with a particular focus on their effectiveness in improving segmentation accuracy and efficiency. This work aims to guide future research and support clinical translation of medical image segmentation, with all related resources publicly available in our GitHub repository: https://github.com/andrew-pengyu/Awsome_MedSeg/tree/main.

04.
arXiv (CS.AI) 2026-06-19

Augmenting Game AI with Deep Reinforcement Learning

arXiv:2606.20210v1 Announce Type: new Abstract: Immersion in video games depends not only on graphics, audio, and game mechanics, but also on the quality of in-game characters. Producing believable characters, or game AI, remains a significant challenge as behavioral complexity is hard to capture with hand-coded systems. Game AI is a source of immersion and engagement; however, the limitations stemming from the challenges of creating game AI often lead to frustration and the breaking of the illusion of realism within the game. The introduction of machine learning models opens the door to creating more believable, authentic, and relatable characters in games. The promise is that they either learn from interacting with the game, or from player data, to develop true human-like behavior. In this paper, we envision more applications of reinforcement learning for game AI in the future. For this to materialize, current research limitations are prohibitive to broad deployment across game genres. Therefore, we propose a framework for training reinforcement learning models with a set of requirements in mind that are suited towards game AI and game development. We present examples of games with reinforcement learning-augmented game AI and describe the practicalities of deploying player-facing machine learning agents in modern games. Furthermore, we identify bottlenecks and hard problems in these areas, which we believe offer promising research directions to accelerate the adoption of machine learning in game AI for the video game industry.

05.
arXiv (CS.AI) 2026-06-24

Cycle-Consistent Neural Explanation of Formal Verification Certificates

arXiv:2606.24414v1 Announce Type: new Abstract: Formal verification produces machine-checkable certificates that attest to the satisfaction or violation of temporal properties, yet these certificates remain opaque to non-specialist stakeholders. We propose a cycle-consistent neural architecture that generates faithful natural language explanations of verification certificates. A forward network NN1 maps certificates to explanations, and an inverse network NN2 reconstructs certificates from explanations; a symbolic verifier closes the loop, providing a differentiable faithfulness proxy. A pointer-generator mechanism ensures lexical grounding by copying state names directly from the certificate. We evaluate on 420 test certificates spanning six verification methods (bounded proof, k-induction, inductive invariant, lasso, reachability, witness pair) in both YES and NO verdict variants, drawn from a financial compliance domain with 207 named states. Our trained architecture, combined with a hybrid inference-time routing strategy, achieves 90.0% cycle-verified soundness, surpassing a multi- LLM few-shot baseline (76.1% for the best of 16 LLM combinations across four frontier models) by 13.9 percentage points. The neural model wins on 10 of 12 verdict/kind categories, with three categories reaching 100% soundness. The architecture offers 860x faster inference (185 ms vs. 160 s per certificate for the full multi-LLM baseline), offline operation, deterministic outputs, and zero per-inference cost. These results demonstrate that trained specialization outperforms general-purpose LLM prompting for structured certificate explanation, while eliminating the deployment constraints of cloud-based inference.

06.
arXiv (CS.LG) 2026-06-17

A fairness-aware extension of Stochastic Multicriteria Acceptability Analysis for ranking

arXiv:2606.17756v1 Announce Type: new Abstract: Fairness has become a central concern in ranking problems involving individuals or social groups, particularly under the Responsible Artificial Intelligence agenda. In Multi-Criteria Decision Analysis, Stochastic Multicriteria Acceptability Analysis (SMAA) provides a robust framework for handling uncertainty and incomplete preference information, but it does not explicitly address fairness in the resulting rankings. This paper proposes SMAA-Fair, a fairness-aware extension of SMAA for ranking problems. The approach reweights the simulated rankings generated by SMAA according to their level of group fairness, so that fairer rankings contribute more strongly to the acceptability indices and central weights vector. The framework is independent of the aggregation model and can incorporate different fairness metrics. In this study, Statistical Parity, normalized discounted Kullback–Leibler divergence (rKL) and normalized discounted cumulative Kullback–Leibler divergence (nDKL) are adopted. Rankings are derived from the fairness-adjusted acceptability matrix using expected ranking and maximum acceptability ranking. We also derive the central weight according to the degree of fairness in the obtained rankings. Numerical experiments with synthetic and real data show that SMAA-Fair improves the representation of protected groups among favourable ranking positions, while preserving robustness to preference uncertainty.

07.
Nature (Science) 2026-06-09

Don’t compete, collaborate: why collective funding applications are the future

Scientists with disparate expertise writing grants together can identify knowledge gaps and drive progress — but systems must change to incentivize them. Scientists with disparate expertise writing grants together can identify knowledge gaps and drive progress — but systems must change to incentivize them.

08.
arXiv (CS.CV) 2026-06-16

MMLongEmbed: Benchmarking Multimodal Embedding Models in Long-Context Scenarios

Recent advancements have significantly expanded the theoretical context windows of Multimodal Embedding Models (MEMs). However, larger context windows do not necessarily translate into effective comprehension and representation of long-context multimodal inputs, which remains a critical bottleneck for real-world deployment. To address the lack of systematic evaluation in this setting, we introduce MMLongEmbed, the first comprehensive benchmark for evaluating MEMs in long-context scenarios. MMLongEmbed comprises four retrieval tasks spanning multiple context-length ranges, covering text, document, and video modalities. Through extensive evaluation of state-of-the-art models, we find that current architectures rely heavily on superficial feature matching and struggle to capture deep semantic and structural dependencies. We further observe that performance degradation varies systematically with context length and key information placement. Moreover, models exhibit substantially different robustness to redundant contextual information across modalities. For reproducibility, the benchmark and code are publicly available.

09.
arXiv (CS.AI) 2026-06-11

Libra: Efficient Resource Management for Agentic RL Post-Training

arXiv:2606.03077v2 Announce Type: replace-cross Abstract: Reinforcement learning (RL) has emerged as a standard post-training paradigm for shaping large language models (LLMs) into capable agents. In agentic RL, the rollout stage generates trajectories while invoking tools, producing long-tailed and non-stationary workloads that expose two fundamental challenges in resource management. First, due to the long-tail distribution, a small fraction of trajectories dominates rollout makespan. Second, rollout and training are subject to cross-stage imbalance, as they exhibit strong asymmetry in compute patterns, memory demands, and sensitivity to sequence length. Compounding this asymmetry, the sequence length distribution drifts continuously as the policy evolves, rendering any static resource split progressively suboptimal. We present Libra, a resource management system to address both challenges via two core mechanisms. The first is a global resource planner that jointly optimizes GPU allocation across rollout and training clusters. It leverages an elastic hybrid pool to enable lightweight, non-blocking worker reallocation between stages. The second is a causality-driven multi-level feedback queue (C-MLFQ) scheduler, which routes requests to heterogeneous rollout buckets based on causal signals derived from tool-return outcomes, rather than relying on fragile length predictions. Evaluated on 48 A800 GPUs, Libra achieves up to 3.0x higher throughput and converges up to 2.5x faster in reward compared to the baselines.

10.
arXiv (quant-ph) 2026-06-24

Semidefinite programming for understanding the limitations of Lindblad equations

arXiv:2602.01794v2 Announce Type: replace Abstract: Lindbladian quantum master equations (LEs) are the most popular descriptions for quantum systems weakly coupled to baths. But, recent works have established that in many situations such Markovian descriptions are fundamentally limited: they cannot simultaneously capture populations and coherences even to the leading-order in system-bath couplings. This can cause violation of fundamental properties like thermalization and continuity equations associated with local conservation laws, even when such properties are expected in the actual setting. This begs the question: given a physical situation, how do we know if there exists an LE that describes it to a desired accuracy? Here we show that, for both equilibrium and non-equilibrium steady states (NESS), this question can be succinctly formulated as a semidefinite program (SDP), a convex optimization technique. If a solution to the SDP can be found to a desired accuracy, then an LE description is possible for the chosen setting. If not, no LE description is fundamentally attainable, showing that a consistent Markovian treatment is impossible even at weak system-bath coupling for that particular setting. Considering few qubit isotropic XXZ-type models coupled to multiple baths, we find that in most parameter regimes, LE description giving accurate populations and coherences to leading-order is unattainable, leading to rigorous no-go results. However, in some cases, LE description having correct populations but inaccurate coherences, and satisfying local conservation laws, is possible over some of the parameter regimes. Our work highlights the power of semidefinite programming in the analysis of physically consistent LEs, thereby, in understanding the limits of Markovian descriptions at weak system-bath couplings.

11.
arXiv (quant-ph) 2026-06-11

Split-Evolution Quantum Phase Estimation for Particle-Conserving Hamiltonians

arXiv:2604.14921v2 Announce Type: replace Abstract: We present a hardware demonstration and resource analysis of split-evolution quantum phase estimation (SE-QPE) on a Quantinuum System Model H2 quantum computer. SE-QPE is a modification to canonical QPE for particle-conserving Hamiltonians in which controlled time evolution is replaced by CSWAP-based interference between a target register and a reference register. For factorizations of time evolution with a shared eigenbasis, SE-QPE preserves the phase-register outcome distribution of canonical QPE and, unlike with compute–uncompute substitutions, it remains compatible with non-exact eigenstates. The substitution removes controlled-simulation overhead and enables parallel evolution on two registers, reducing the depth of each phase-kickback block. Resource analysis for Trotterized double-factorized chemistry Hamiltonians shows that the substitution becomes increasingly favorable at higher phase powers and combining QPE and SE-QPE implementations can be a useful option. Over a range of FeMoco active spaces, SE-QPE reduces time evolution resources, with asymptotic reductions of about 33% in CX count, 25% in $T$ count, and an asymptotic depth ratio of $3/N$ for CX layers. On Quantinuum H2-2, a four-qubit model ethylene demonstration with explicit inverse QFT and repeated phase-kickback steps up to 8 phase bits yields distinct energies and shows the auxiliary registers provide useful error detection filters.

12.
arXiv (quant-ph) 2026-06-12

Theoretical Study for Generating Optical GKP State via a Single-Photon-Added Squeezed Vacuum

arXiv:2606.12467v1 Announce Type: new Abstract: A theoretical framework is developed to analyze the generation of the optical GKP state using a single-photon-added squeezed vacuum. This state, defined by the squeezing parameter $r$, is injected into a 50:50 beam splitter, and the optical GKP state is obtained through conditional measurement at one output port. The single-photon-added squeezed vacuum is especially prominent in this context because it provides a simpler and more experimentally accessible ingredient than Schrodinger cat states, while conditional measurement ensures projection onto a state that closely approximates the finite-energy GKP form. Fidelity is employed to quantify this closeness, and the analysis demonstrates that the scheme achieves a maximum fidelity of 85% at a squeezing level of $3.76 \ dB$. This performance surpasses approaches based on squeezed optical odd Schrodinger cat states, underscoring the single-photon-added squeezed vacuum as a practical and effective pathway toward fault-tolerant photonic quantum computing.

13.
arXiv (CS.AI) 2026-06-11

LLMs+Graphs: Toward Graph-Native, Synergistic AI Systems

arXiv:2606.11560v1 Announce Type: cross Abstract: Large Language Models (LLMs) have advanced rapidly, but their limitations in structured and multi-hop reasoning underscore the need for graph-native, synergistic artificial intelligence (AI) systems. Graph-structured data underpins critical applications across social, biological, financial, transportation, web, and knowledge domains, making it essential to understand how LLMs can leverage graph computation for grounded, context-rich inference. Three complementary synergies are emerging: LLMs augmented with graph computation for retrieval and reasoning; bidirectional integration between LLMs and knowledge graphs (KGs), where LLMs support KG construction and curation while KGs enforce semantic constraints and factual consistency; and AI agents strengthened by graph algorithms for planning, decision making, and multi-step reasoning. In parallel, LLMs introduce new capabilities for graph data management and graph machine learning (ML) through natural language interfaces and hybrid LLM-graph neural network (GNN) pipelines. This tutorial synthesizes the algorithms, systems, and design principles driving these converging directions, offering data science and data mining researchers a unified perspective on integrating LLMs, graph data management, graph mining, graph ML, and agentic computation into next-generation graph-native AI systems.

14.
arXiv (CS.CV) 2026-06-15

Avatar V: Scaling Video-Reference Avatar Video Generation

Generating avatar videos that are not merely visually similar to a target individual but behaviorally recognizable, faithfully reproducing their talking rhythm, gestural tendencies, and expression dynamics, remains an open challenge. Existing methods predominantly condition on single static images, which provide insufficient identity information and cannot capture dynamic motion traits, while standard pixel-level objectives underserve the perceptually critical facial regions that determine avatar fidelity. We present Avatar V, a production-scale framework that addresses these limitations through video-reference-conditioned identity modeling. Rather than compressing identity into fixed-size embeddings, the model conditions directly on the full token sequence of a reference video, learning to reproduce both static identity attributes (facial geometry, skin texture) and dynamic behavioral patterns (talking rhythm, micro-expressions) through attention over the reference context. We introduce Sparse Reference Attention, an asymmetric mechanism achieving linear-complexity conditioning on arbitrarily long references; a motion representation stream enabling closed-loop talking style transfer; and an identity-aware super-resolution refiner inheriting the full reference conditioning. These are supported by a data engine curating 100M+ training clips from 50M raw videos, and a five-stage training pipeline with flow matching pre-training, personality fine-tuning, two-phase distillation (>10x acceleration), and RLHF alignment, deployed across thousands of GPUs. Avatar V generates 1080p videos of unlimited duration, achieving state-of-the-art identity preservation, lip synchronization, and generation quality on our cross-scene benchmark, consistently outperforming leading systems including Seedance 2.0, Kling O3 Pro, Veo 3.1, and OmniHuman 1.5 in both automated metrics and human evaluation.

15.
arXiv (CS.LG) 2026-06-24

Relatively Smart: A New Approach for Instance-Optimal Learning

arXiv:2603.01346v2 Announce Type: replace Abstract: We revisit the framework of Smart PAC learning, which seeks supervised learners which compete with semi-supervised learners that are provided full knowledge of the marginal distribution on unlabeled data. Prior work has shown that such marginal-by-marginal guarantees are possible for "most" marginals, with respect to an arbitrary fixed and known measure, but not more generally. We discover that this failure can be attributed to an "indistinguishability" phenomenon: There are marginals which cannot be statistically distinguished from other marginals that require different learning approaches. In such settings, semi-supervised learning cannot certify its guarantees from unlabeled data, rendering them arguably non-actionable. We propose relatively smart learning, a new framework which demands that a supervised learner compete only with the best "certifiable" semi-supervised guarantee. We show that such modest relaxation suffices to bypass the impossibility results from prior work. In the distribution-free setting, we show that the One-Inclusion Graph learner is relatively smart up to squaring the sample complexity, and show that no supervised learning algorithm can do better. For distribution-family settings, we show that relatively smart learning can be impossible or can require idiosyncratic learning approaches, and its difficulty can be non-monotone in the inclusion order on distribution families.

17.
arXiv (CS.CV) 2026-06-19

MMD-SLAM: Structure-Enhanced Multi-Meta Gaussian Distribution-Guided Visual SLAM

3D Gaussian Splatting (3DGS) has significantly boosted novel view synthesis and high-fidelity scene reconstruction, expanding the potential of 3DGS-based Visual Simultaneous Localization and Mapping (SLAM) methods. However, most existing systems fail to fully exploit the underlying structural information, which limits rendering quality and often leads to inconsistent maps. To address these limitations, we propose MMD-SLAM, a structure-enhanced Visual SLAM framework that leverages the Atlanta World (AW) assumption to guide a Multi-Meta Gaussian representation for photorealistic mapping. First, we introduce a point-line fusion strategy for pose optimization, where 3D line segments are incorporated to improve tracking robustness and provide additional constraints for mapping. Second, we design a Multi-Meta Gaussian representation with dominant directions, explicitly encoding structural priors from the AW hypothesis. Finally, we propose a Gaussian evolution strategy that adapts to scene geometry and incorporates structural cues into global optimization. Extensive experiments demonstrate that these innovations enable MMD-SLAM to achieve state-of-the-art performance in both tracking accuracy and mapping quality. e.g., our method achieves a 48.56% reduction in ATE RMSE on ScanNet and a 5.71% improvement in PSNR on Replica, compared with MonoGS.

18.
PLOS Medicine 2026-05-06

Point-of-care early infant HIV diagnosis at birth in a pragmatic cluster-randomized trial in Mozambique and Tanzania: A comparative cost and cost-effectiveness study

by Kira Elsbernd, Issa Sabi, Ilesh V. Jani, Chishamiso Mudenyanga, Siriel Boniface, Arlete Mahumane, Joaquim Lequechane, Falume Chale, Bindiya Meggi, Kassia Pereira, Raphael Edom, Anange F. Lwilla, W. Chris Buck, Nyanda Elias Ntinyinya, Michael Hoelscher, Till Baernighausen, Arne Kroidl, Stefan Kohler, the LIFE Study Consortium Background Timely access to early infant diagnosis (EID) is crucial for newborns with HIV, as late diagnosis can delay lifesaving antiretroviral treatment (ART). We assessed the comparative cost and cost-effectiveness of integrating point-of-care EID at birth into routine care in primary healthcare settings. Methods and findings This pre-specified secondary analysis was nested in the cluster-randomized LIFE study conducted at 28 primary healthcare facilities in Mozambique and Tanzania from October 2019 to September 2021. We estimated the health system cost of point-of-care birth plus 4–8-week HIV testing (very early infant diagnosis; VEID) compared to standard-of-care (SoC) testing at 4–8 weeks only, both with immediate ART initiation. We assessed the cost-effectiveness of VEID relative to SoC with respect to ART initiation within one week of life using Bayesian hierarchical models. As this is an intermediate outcome, incremental cost-effectiveness ratios (ICERs) cannot be directly compared to available life-year-based cost-effectiveness thresholds. To contextualize results, we derived the minimum life-years gained per early ART initiation required for VEID to meet standard thresholds in a break-even analysis.VEID was associated with a higher cost and resulted in earlier ART initiation than SoC in both countries. In Mozambique, VEID increased the proportion of infants initiating ART within one week of life by 90.0 (95% CrI [67.5, 98.5]) percentage points at an incremental cost of $2,632 (95% CrI [$2,249, $3,062]) per infant with HIV. In Tanzania, VEID increased early ART initiation by 59.9 (95% CrI [20.9, 89.5]) percentage points at an incremental cost of $6,263 (95% CrI [$5,394, $7,243]) per infant with HIV. The ICER was $2,924 and $10,458 in Mozambique and Tanzania, respectively and was sensitive to intrauterine transmission rate. These findings were limited by the lack of long-term health outcome data and reliance on an intermediate outcome. Based on the break-even analysis, we estimated that VEID would need to yield 6–32 life-years gained per additional early ART initiation to meet standard thresholds. Conclusions Adding birth testing improved early ART initiation but was unlikely to be cost-effective relative to standard thresholds given current prices, vertical transmission rates, and knowledge of long-term health benefits. Cost-effectiveness could be achieved at current costs if early ART translates to substantial long-term health benefits or if targeted to infants at high risk of vertical transmission.

19.
medRxiv (Medicine) 2026-06-17

Deep learning for interactive and automated inner retinal layer segmentation in OCT images of patients with retinitis pigmentosa using limited training data

Purpose: New therapeutic strategies such as optogenetics have created a need for accurate tracking of inner retina degeneration in Retinitis pigmentosa (RP) patients. We introduce two tailored deep learning models to segment the RNFL (retinal nerve fibre layer), GCIPL (ganglion cell inner plexiform layer), INL (inner nuclear layer), CFT (central foveal thickness) and RPE (retinal pigment epithelium) in RP: The first is based on a Segment Anything Model (SAM), the second on nnU-Net. To our knowledge, SAM has not yet been applied to retinal layers in OCT data. Methods: SD-OCT images of a retrospective cohort of 37 RP patients were included. Data for four training cycles were prepared semi-automatically in MATLAB, then assessed and corrected by three expert graders. 1,700 segmented B-Scans from two open datasets were used for pretraining. For post-processing, semantic retinal boundary detection was developed. The final models, OCT-SAM and nnU-Net, were trained on 228 annotated RP scans. Detected layer thicknesses were validated against manual segmentation at 90 random points in 30 OCT B-Scans. Finally, OCT-SAM was tested on three RP cases with retrospective, longitudinal OCT data. Results: nnU-Net achieved a precision, recall and F-1 score of 0.96 while OCT-SAM performance resulted in slightly lower values of 0.93, 0.8 and 0.85, respectively. OCT-SAM measurements had low bias and good agreement with manual annotations, confirming reliability. Conclusions: OCT-SAM enabled fast data annotation and tool integration, whereas nnU-Net provided the best segmentation performance. OCT-SAM demonstrated longitudinal reproducibility and detected RP-characteristic pathologies and degenerative changes. Future work will extend OCT-SAM to 3D OCT segmentation.

20.
arXiv (CS.CL) 2026-06-19

Quality Over Clicks: Iterative Reinforcement Learning for Early-Stage E-Commerce Query Suggestion

Existing dialogue systems rely on query suggestion to enhance user engagement. Recent approaches mainly optimize generative models using click-through rate (CTR) models to align with user preferences. However, these methods are less effective in early-stage deployment scenarios, where click feedback is sparse and insufficient for training a reliable CTR model. To bridge this gap, we propose QualEQS, a quality-first iterative reinforcement learning framework for e-commerce query suggestion. We formalize actionable suggestion quality along three dimensions that directly affect downstream usability: answerability, factuality, and information gain. To continuously improve from online traffic without click supervision, we further propose group-level disagreement among candidate suggestions to identify ambiguous query contexts and mine hard training cases for iterative refinement. We also introduce EQS-Benchmark, a dataset of 16,949 real-world e-commerce queries for offline training and evaluation. Experiments show that our quality-based offline metrics correlate strongly with online performance, providing a practical evaluation recipe for sparse-feedback deployment. In both offline and online settings, QualEQS consistently outperforms strong baselines, yielding a 6.81% improvement in online ChatPV in a real-world enterprise-level conversational shopping assistant system.

21.
arXiv (CS.AI) 2026-06-16

Metric Match: A Subset Selection Approach to Evaluating LLM Judge Reliability

arXiv:2606.15029v1 Announce Type: new Abstract: LLM judges are used to reduce the need for costly human labor in evaluating open-ended text generation. However, the reliability of these judges depends critically on their alignment with human raters – a property that itself depends on costly human annotations. In this work, we develop a method (Metric Match) for estimating correlation-based reliability metrics of LLM judges from limited annotations. Metric Match selects a subset of samples for human annotation such that the subset matches the population reliability metric with respect to acquired synthetic labels. We empirically show that Metric Match achieves a win-rate of 0.838 against random subset selection across four different correlation metrics and 15 datasets, with an 18.7% decrease in average estimation error and reduces annotation needs by 32.5%. We provide a cost model and highlight a medical case study where our method saves $1,041.67 compared to random selection for expert annotation. Further, we shift our task from reliability estimation to reliability classification of whether a given judge is above a deployment threshold, outperforming random selection with Metric Match. All project code is publicly available, and we additionally provide an installable package for ease of use.

22.
arXiv (CS.CV) 2026-06-18

RUB: Evaluating Residual Knowledge in Unlearned Models

Machine Unlearning (MUL) has emerged as a key mechanism for privacy protection and content regulation, yet current techniques often fail to guarantee the complete removal of sensitive information. While most existing works focus on verifying the execution of unlearning, they overlook the critical question of whether models remain robust against adversarial attempts to recover forgotten knowledge. In this work, we advocate for the principle of Robust Unlearning, which requires models to be both indistinguishable from retrained counterparts and resilient against diverse adversarial threats. To instantiate this principle, we propose a unified benchmark, RUB (Robust Unlearning Benchmark), that systematically evaluates the robustness of unlearning algorithms across classification, image-to-image reconstruction, and text-to-image synthesis. Within this framework, we introduce the Unlearning Mapping Attack (UMA) as a generalizable method to detect residual information, and demonstrate how existing attack strategies can be adapted into this framework as long as they conform to the generic UMA framework. Our experiments across discriminative and generative tasks reveal that state-of-the-art unlearning methods remain vulnerable under these evaluations, even when passing standard verification metrics. By positioning robustness as the central criterion and providing a benchmark for adversarial evaluation, we hope RUB paves the way toward more reliable and secure unlearning practices. The codebase and model checkpoints in RUB will be published.

23.
arXiv (CS.CL) 2026-06-11

GrowLoop: Self-Evolving Conversation Evaluation Seeded by Human

With the rapid advancement of large language models, evaluating human-likeness in open-ended conversation has become increasingly important. However, human-likeness is a form of tacit knowledge that humans perceive intuitively, yet the underlying criteria resist explicit formulation. Human judgments vary widely, with strong agreement on some cases and legitimate disagreement on others. Meanwhile, the criteria behind human judgments remain implicit, leaving no clear basis for constructing cases. Further, what counts as human-likeness is not static, but evolving with model capability and human expectations. Despite progress in evaluation methods such as expert-authored benchmarks, Reward Models, and self-evolving benchmarks, none addresses all three challenges simultaneously. Therefore, we propose GrowLoop, a self-evolving conversation evaluation system that continuously adapts as models advance and scenarios shift. Starting from minimal human seed annotations, LLM agents iteratively extract and refine evaluation rubrics through Heuristic Learning. Human-AI agreement is required where annotators converge, while only plausibility is expected where they diverge. Moreover, the Rubric-Case co-evolution mechanism enables continuous evolution. When the evaluation target shifts, new human seeds expand the system's coverage accordingly. When applied to human-likeness evaluation in open-ended conversation, the AI judge guided by these rubrics not only substantially outperforms existing methods in alignment with human judgments, but also uncovers issues that annotators overlook. The resulting benchmark effectively discriminates models across capability tiers and reveals where they fall short, while generalizing to new scenarios and adapting as models advance. Our work shifts the benchmarking paradigm from manual updates or difficulty scaling to comprehensive, continuous self-evolution.

24.
arXiv (CS.AI) 2026-06-15

The Weight Norm Sets the Grokking Timescale: A Causal Delay Law

arXiv:2606.13753v1 Announce Type: cross Abstract: Grokking is the delayed onset of generalization in neural networks, arising long after they fit the training data. Whether the weight norm causes this delay is disputed: some studies report a critical norm at the transition, others observe grokking with no fixed norm at all. We settle this by intervening on the norm during training rather than only observing it. Under free training with weight decay, networks grok when the weight norm reaches a value Wc that varies little across seeds and learning rates (CV 1 to 2 percent) and grows with the modular base as a power law. When we instead clamp the norm to a fixed multiple rho of Wc and hold it there, the network still groks, but the delay follows T_grok proportional to exp(alpha rho). One exponent, alpha near 7.5, fits this delay across four moduli (R^2 = 0.996). Over the swept ranges the held norm moves the delay by about 19x and the learning rate by only about 2x, and holding the norm above Wc slows grokking rather than preventing it. A final LayerNorm removes the dependence by decoupling weight scale from the network function; without it the exponential law returns. This pinned-norm delay is the exponential counterpart to the logarithmic delay predicted for a freely contracting norm.

25.
arXiv (CS.LG) 2026-06-15

Beyond a Single Explanation of the Adam–SGD Gap

arXiv:2606.14259v1 Announce Type: new Abstract: Prior work has identified several factors that can contribute to the performance gap between Adam and SGD, spanning data aspects, architecture design, and optimization properties. Yet these explanations are often studied in isolation, leaving their relative importance unclear. In this work, we revisit these hypotheses through a controlled empirical study across vision, language, genomics, and graph tasks, spanning modern and classical architectures, and carefully designed training setups. Our results suggest that no single factor consistently explains the Adam–SGD gap. For instance, the Adam advantage can (1) persist under a uniform vocabulary distribution yet nearly disappear under a heavy-tailed one; (2) reverse in favor of SGD in softmax-attention models; and (3) become larger under soft architectural modifications, e.g., when ReLU is replaced by a GeLU nonlinearity. This suggests that the gap arises from nontrivial data and architecture interactions, rather than from a single common factor. Yet, we observe a pattern across our settings: a crossover batch size at which the relative advantage shifts from SGD to Adam as the batch size scales. These empirical results are captured by our theoretical gap model, which predicts this batch-size-dependent crossover. Our perspective helps reconcile several existing hypotheses while offering practical insights across domains.