Automatic evaluation with large language models, commonly known as LLM-as-a-judge, is now standard across reasoning and alignment tasks. Despite evaluating many samples in deployment, these evaluators typically (i) treat each case independently, missing the opportunity to accumulate experience, and (ii) rely on a single fixed prompt for all cases, neglecting the need for sample-specific evaluation criteria. We introduce Learning While Evaluating (LWE), a framework that allows evaluators to improve sequentially at inference time without requiring training or validation sets. LWE maintains an evolving meta-prompt that (i) produces sample-specific evaluation instructions and (ii) refines itself through self-generated feedback. Furthermore, we propose Selective LWE, which updates the meta-prompt only on self-inconsistent cases, focusing computation where it matters most. This selective approach retains the benefits of sequential learning while being far more cost-effective. Across two pairwise comparison benchmarks, Selective LWE outperforms strong baselines, empirically demonstrating that evaluators can improve during sequential testing with a simple selective update, learning most from the cases they struggle with.

Open-vocabulary mobile manipulation (OVMM) requires robots to follow language instructions, navigate, and manipulate while updating their world representation as the environment changes dynamically. However, most prior works update their world representation only at discrete milestones, such as waypoints or the end of an action step. Such sparse updates leave robots with limited awareness between updates, causing missed objects, delayed error detection, and slower replanning. To address this limitation, we propose BINDER (Bridging INstant and DEliberative Reasoning), a dual-process framework that separates strategic planning from continuous environmental monitoring. BINDER combines a Deliberative Response Module (DRM, a multimodal LLM for task planning) with an Instant Response Module (IRM, a Video-LLM for continuous monitoring). The DRM handles strategic planning through structured 3D scene updates and guides the IRM’s focus, while the IRM processes video streams to update memory, proactively adjust actions, and trigger replanning when needed. This bidirectional coordination ensures continuous awareness without costly updates, enabling reliable and robust operation under dynamic conditions. We evaluate BINDER in three real-world environments where objects are moved during execution and show that it achieves substantially higher success rates and efficiency than state-of-the-art baselines, confirming its effectiveness for real-world deployment.

Vision-Language-Action (VLA) models have emerged as a promising paradigm for general-purpose robotic control, with test-time scaling (TTS) gaining attention to enhance robustness beyond training. However, existing TTS methods for VLAs require additional training, verifiers, and multiple forward passes, making them impractical for deployment. Moreover, they intervene only at action decoding while keeping visual representations fixed—insufficient under perceptual ambiguity, where reconsidering how to perceive is as important as deciding what to do. To address these limitations, we propose SCALE, a simple inference strategy that jointly modulates visual perception and action based on 'self-uncertainty', inspired by uncertainty-driven exploration in Active Inference theory—requiring no additional training, no verifier, and only a single forward pass. SCALE broadens exploration in both perception and action under high uncertainty, while focusing on exploitation when confident—enabling adaptive execution across varying conditions. Experiments on simulated and real-world benchmarks demonstrate that \method improves state-of-the-art VLAs and outperforms existing TTS methods while maintaining single-pass efficiency.

Video Large Multimodal Models (VLMMs) have shown impressive performance in video understanding, yet their ability to accurately capture the temporal order of multiple events remains underexplored. We interestingly observe that, even when video frames are scrambled, models perform very well on the existing benchmarks by comprehensive experiments. This implies that VLMMs may not necessarily rely on accurate sequential processing of visual events, but instead depend on prior knowledge of typical scenarios to answer the question. To benchmark temporal understanding capabilities in VLMMs, we propose VECTOR, designed to explicitly assess a model's ability to identify the temporal order of events. On this benchmark, we observe that various VLMMs often fail to understand the orders of events. To address this, we propose MECOT (Multi-Event instruction fine-tuning with Chain-of-Thought), which (1) trains models on detailed, event-by-event video descriptions and (2) using chain-of-thought prompts at inference to enhance temporal awareness. MECOT outperforms prior arts on VECTOR as well as improving performance on existing video benchmarks, implying effectiveness of temporal understanding. We release our code, model and datasets.

Modern Vision-Language Models (VLMs) often struggle with strategic reasoning, i.e., anticipating and influencing other agents' actions, under uncertainty in competitive and cooperative settings. Real-time strategy (RTS) games can be a natural testbed for diagnosing this limitation, as they demand coordination with allies, adaptation to opponents' strategy, and long-horizon planning under partial observability. However, existing RTS benchmarks offer limited evaluation scope, lack systematic competency diagnosis, and remain fixed in the pre-designed scenario coverage. To address these limitations, we present RTSGameBench, which is built on Beyond All Reason, a large-scale RTS game with an expanded battlefield that demands broader strategy diversity than the existing testbeds. The proposed benchmark provides evaluations through diverse gameplay across various matchup structures, diagnostic assessment via mini-games, each targeting an individual strategic competency, and extensible coverage via a self-evolving generation framework that converts free-form queries into new mini-games, improving over successive cycles. Additionally, for VLMs to operate in large-scale RTS games, we provide RTSGameAgent that manages units by an FSM with agentic memory. We empirically validate that multiple state-of-the-art VLMs do not perform well when matchups demand tighter coordination, multiagent coordination and when task scale increases.

Iterative self-improvement, a concept extending beyond personal growth, has found powerful applications in machine learning, particularly in transforming weak models into strong ones. While recent advances in natural language processing have shown its efficacy through iterative preference optimization, applying this approach to Video Large Multi-modal Models (VLMMs) remains challenging due to modality misalignment. VLMMs struggle with this misalignment during iterative preference modeling, as the self-judge model often prioritizes linguistic knowledge over visual information. Additionally, iterative preference optimization can lead to visually hallucinated verbose responses due to length bias within the self-rewarding cycle. To address these issues, we propose Iterative Self-Retrospective Direct Preference Optimization (ISR-DPO), a method that uses self-retrospection to enhance preference modeling. This approach enhances the self-judge's focus on informative video regions, resulting in more visually grounded preferences. In extensive empirical evaluations across diverse video question answering benchmarks, the ISR-DPO significantly outperforms the state of the art. We are committed to open-sourcing our code, models, and datasets to encourage further investigation.

Video corpus moment retrieval (VCMR) task aims to retrieve a specific moment from a large corpus of untrimmed videos. This task has been addressed by decomposing it into video retrieval and moment retrieval subtasks each with specialized heads (i.e., democratic decomposition), due to the computational complexity. However, this approach overlooks the interdependency between the subtasks, which is crucial for temporally fine-grained query-video alignment. To address this suboptimality, we propose an integrated learning framework that explicitly establishes a connection between the two subtasks, namely moment-aware video retrieval, allowing them to benefit from each other’s learning process. Furthermore, we employ a curriculum-based negative sampling strategy that gradually provides harder negative samples to enhance the discriminative ability between semantically similar negative videos. We empirically show that the proposed approach outperforms state-of-the-art methods on three benchmarks – TVR, ActivityNet, and DiDeMo. Notably, on TVR, our method achieves 10.04% inVCMRR@1with tIoU=0.7, representing a 1.68% absolute improvement over prior work, and similarly demonstrates consistent gains on ActivityNet (4.98% in VCMR R@1 with tIoU=0.5) and DiDeMo, validating the effectiveness of our integrated approach.

Large Language Models (LLMs) have recently demonstrated impressive action sequence prediction capabilities but often struggle with dynamic, long-horizon tasks such as real-time strategic games. In a game such as StarCraftII (SC2), agents need to manage resource constraints and adapt to evolving battlefield situations in a partially observable environment. This often overwhelms exisiting LLM-based approaches. To address these challenges, we propose a hierarchical multi-agent framework that employs specialized imitation learning agents under a meta-controller called Strategic Planner (SP). By expert demonstrations, each specialized agent learns a distinctive strategy, such as aerial support or defensive maneuvers, and produces coherent, structured multistep action sequences. The SP then orchestrates these proposals into a single, environmentally adaptive plan that ensures local decisions aligning with long-term strategies. We call this HIMA (Hierarchical Imitation Multi-Agent). We also present TEXTSCII-ALL, a comprehensive SC2 testbed that encompasses all race match combinations in SC2. Our empirical results show that HIMA outperforms state of the arts in strategic clarity, adaptability, and computational efficiency, underscoring the potential of combining specialized imitation modules with meta-level orchestration to develop more robust, general-purpose AI agents.

Recent advancements in large language models have influenced the development of video large multimodal models (VLMMs). The previous approaches for VLMMs involved Supervised Fine-Tuning (SFT) with instruction-tuned datasets, integrating LLM with visual encoders, and adding additional learnable modules. Video and text multimodal alignment remains challenging, primarily due to the deficient volume and quality of multimodal instruction-tune data compared to text-only data. We present a novel alignment strategy that employs multimodal AI system to oversee itself called Reinforcement Learning from AI Feedback (RLAIF), providing self-preference feedback to refine itself and facilitating the alignment of video and text modalities. In specific, we propose context-aware reward modeling by providing detailed video descriptions as context during the generation of preference feedback in order to enrich the understanding of video content. Demonstrating enhanced performance across diverse video benchmarks, our multimodal RLAIF approach, VLM-RLAIF, outperforms existing approaches, including the SFT model. We commit to open-sourcing our code, models, and datasets to foster further research in this area.

Story visualization (SV) is a challenging text-to-image generation task for the difficulty of not only rendering visual details from the text descriptions but also encoding a longterm context across multiple sentences. While prior efforts mostly focus on generating a semantically relevant image for each sentence, encoding a context spread across the given paragraph to generate contextually convincing images (e.g., with a correct character or with a proper background of the scene) remains a challenge. To this end, we propose a novel memory architecture for the Bi-directional Transformer framework with an online text augmentation that generates multiple pseudo-descriptions as supplementary supervision during training for better generalization to the language variation at inference. In extensive experiments on the two popular SV benchmarks, i.e., the Pororo-SV and Flintstones-SV, the proposed method significantly outperforms the state of the arts in various metrics including FID, character F1, frame accuracy, BLEU-2/3, and R-precision with similar or less computational complexity.

최근 자연어 기반 비디오 탐색 알고리즘 연구들은 대부분 이미 존재하는 레이블들을 활용한 데이터셋을 바탕으로 완전지도학습 혹은 준지도학습의 알고리즘들을 기반으로 하고 있다. 그러나 이러한 데이터셋의 구축에는 많은 비용이 들어가며, 레이블을 만들기 어려운 현실 세계에서 사용하기 적합하지 않다. 그렇기에 본 연구에서는 외부지식을 활용한 자연어 기반 비디오 탐색 알고리즘(EK-NLVL)을 제안하며, 사전 학습된 캡셔닝 모델과 비지도 기반의 비디오 영역 탐색 기법을 통해 효과적인 pseudo-supervision을 모델에 줄 수 있는 프레임워크를 제안한다. 거기에 더해 대규모 데이터셋에 사전 학습된 멀티 모달 표현 학습 모델인 CLIP을 활용하여 기존의 자연어 증강 기법인 역번역기법을 바탕으로 시각 정보와 텍스트 정보를 동기화 시켜 pseudo-sentence의 정보의 품질을 향상 시키는 Visual-Aligned Sentence Filtering(VAF) 데이터 필터링 기법을 제안한다. 이렇게 외부지식을 통해 생성된 데이터를 효과적으로 활용할 수 있는 Query-Attentive on Segmentation(QAS) 모델 또한 제안하며 Charades-STA 데이터셋에서의 실험을 통해 EK-NLVL 방법론의 효과를 볼 수 있다.

Understanding videos to localize moments with natural language often requires large expensive annotated video regions paired with language queries. To eliminate the annotation costs, we make a first attempt to train a natural language video localization model in zero-shot manner. Inspired by unsupervised image captioning setup, we merely require random text corpora, unlabeled video collections, and an off-the-shelf object detector to train a model. With the unpaired data, we propose to generate pseudo-supervision of candidate temporal regions and corresponding query sentences, and develop a simple NLVL model to train with the pseudo-supervision. Our empirical validations show that the proposed pseudo-supervised method outperforms several baseline approaches and a number of methods using stronger supervision on Charades-STA and ActivityNet-Captions.

An erasing method capable of speeds 10(4)-fold faster compared with those by the conventional Fowler-Nordheim (FN) erasing technique is experimentally demonstrated in a gate-all-around junction less (JL) charge-trap flash memory device using thermal excitation with the aid of electric field. A gate electrode serving as a built-in heater generates Joule heat for thermal excitation of trapped electrons in the charge-trap layer. The electrons excited by thermal excitation are further accelerated for injection into the silicon body by an in-situ applied E-field between the gate and the JL body. Hence, the trapped electrons are removed within 70 ns by electro-thermal erasing.

A silicon-nanowire (Si-NW) switch with triboelectricity is proposed for the first time for future electronics with enhanced security application. The dimensions of the mechanical NW switch are width 50 nm and thickness 100 nm with a 50-nm-thick airgap, fabricated with aid of CMOS fabrication technology. By hand-touch on gate pad, the fabricated switch can turn-on as well as turn-off by destruction of fin-channel which comes from electrostatic force by triboelectricity. This technology is expected as a security-enhanced future electronics.

This paper demonstrates a breakthrough for DRAM scaling: A vertically integrated gate-all-around (GAA) silicon nanowire (SiNW) channel-based dynamic random access memory (DRAM) without a cell capacitor for data storage, i.e., a zero-capacitor DRAM unlike the conventional DRAM. Vertical integration of the SiNW was attained by a one-route all-dry etching process (ORADEP), resulting in stiction-free stability and simplicity in the fabrication process. High performance that is suitable for high packing density integration is presented with vertically integrated multiple channels, which reveals a potential for an ultimate scaling of DRAM toward the end of the roadmap. (C) The Author(s) 2016.

We report an accurate extraction technique of effective mobility (mu(eff)) by considering gate-bias (V-GS) dependent effective inversion charges (Q(inv,eff)) as a normalized correction factor (sigma(V-GS)) when using the gate-to-source/drain capacitance-voltage (C-V) measurement in a p-channel Si/SiGe MOSFET with a floating body structure. In the proposed technique, two different capacitance-voltage configurations: a gate-to-source/drain (CG-SD) configuration without body contact and a gate-to-source/drain/body (CG-SDB) configuration with body contact are utilized for the accurate extraction of mu(eff).

An ultra-fast erasing process that acts within 200 ns is demonstrated in a junctionless gate-all-around nanowire silicon-oxide-nitride-oxide-silicon device. Rapid erasing is enabled with the use of instantaneous thermal excitation (TE) through a double-ended gate structure. Charges inside the silicon nitride layer are de-trapped by Joule heating. Moreover, an in-situ self-annealing effect accompanied by the TE erase method is achieved; hence, both the tunnel oxide quality and the retention characteristics are less degraded compared with the conventional Fowler-Nordheim erase method.

The influence of process-induced defect formation was investigated in a vertically integrated (VI) junctionless-mode field-effect transistor (JL-FET). Compared to the low energy and one-time ion-implantation process to fabricate a single nanowire-based FET, the high-energy and repetitive ion-implantation process for the creation of the VI JL-FET inevitably generates more defects in the crystalline sites. Even after high-temperature rapid thermal annealing, the non-recovered defect sites existing in the interface and silicon channel, as verified by a transmission electron microscopy analysis, lead to the degradation of the electrical performance such as on-and off-state current. Particularly, the abnormal behavior of the off-state current, mostly arising from the gate-induced drain leakage, was analyzed using the experimental results, and supported by the numerical simulation as well. Published by AIP Publishing.