❌

Normal view

Received — 17 March 2026 ⏭

Diagnosing Retrieval Bias Under Multiple In-Context Knowledge Updates in Large Language Models

✇cs.AI, q-bio.NC updates on arXiv.org
By: Boyu Qiao · Sean Guo · Xian Yang · Kun Li · Wei Zhou · Songlin Hu · Yunya Song
16 March 2026 at 12:00
arXiv:2603.12271v1 Announce Type: cross Abstract: LLMs are widely used in knowledge-intensive tasks where the same fact may be revised multiple times within context. Unlike prior work focusing on one-shot updates or single conflicts, multi-update scenarios contain multiple historically valid versions that compete at retrieval, yet remain underexplored. This challenge resembles the AB-AC interference paradigm in cognitive psychology: when the same cue A is successively associated with B and C, the old and new associations compete during retrieval, leading to bias. Inspired by this, we introduce a Dynamic Knowledge Instance (DKI) evaluation framework, modeling multi-updates of the same fact as a cue paired with a sequence of updated values, and assess models via endpoint probing of the earliest (initial) and latest (current) states. Across diverse LLMs, we observe that retrieval bias intensifies as updates increase, earliest-state accuracy stays high while latest-state accuracy drops substantially. Diagnostic analyses of attention, hidden-state similarity, and output logits further reveal that these signals become flatter and weakly discriminative on errors, providing little stable basis for identifying the latest update. Finally, cognitively inspired heuristic intervention strategies yield only modest gains and do not eliminate the bias. Our results reveal a persistent challenge in tracking and following knowledge updates in long contexts.

SRAM-Based Compute-in-Memory Accelerator for Linear-decay Spiking Neural Networks

✇cs.AI, q-bio.NC updates on arXiv.org
By: Hongyang Shang · Shuai Dong · Yahan Yang · Junyi Yang · Peng Zhou · Arindam Basu
16 March 2026 at 12:00
arXiv:2603.12739v1 Announce Type: cross Abstract: Spiking Neural Networks (SNNs) have emerged as a biologically inspired alternative to conventional deep networks, offering event-driven and energy-efficient computation. However, their throughput remains constrained by the serial update of neuron membrane states. While many hardware accelerators and Compute-in-Memory (CIM) architectures efficiently parallelize the synaptic operation (W x I) achieving O(1) complexity for matrix-vector multiplication, the subsequent state update step still requires O(N) time to refresh all neuron membrane potentials. This mismatch makes state update the dominant latency and energy bottleneck in SNN inference. To address this challenge, we propose an SRAM-based CIM for SNN with Linear Decay Leaky Integrate-and-Fire (LD-LIF) Neuron that co-optimizes algorithm and hardware. At the algorithmic level, we replace the conventional exponential membrane decay with a linear decay approximation, converting costly multiplications into simple additions while accuracy drops only around 1%. At the architectural level, we introduce an in-memory parallel update scheme that performs in-place decay directly within the SRAM array, eliminating the need for global sequential updates. Evaluated on benchmark SNN workloads, the proposed method achieves a 1.1 x to 16.7 x reduction of SOP energy consumption, while providing 15.9 x to 69 x more energy efficiency, with negligible accuracy loss relative to original decay models. This work highlights that beyond accelerating the (W x I) computation, optimizing state-update dynamics within CIM architectures is essential for scalable, low-power, and real-time neuromorphic processing.

SortScrews: A Dataset and Baseline for Real-time Screw Classification

✇cs.AI, q-bio.NC updates on arXiv.org
By: Tianhao Fu · Bingxuan Yang · Juncheng Guo · Shrena Sribalan · Yucheng Chen
16 March 2026 at 12:00
arXiv:2603.13027v1 Announce Type: cross Abstract: Automatic identification of screw types is important for industrial automation, robotics, and inventory management. However, publicly available datasets for screw classification are scarce, particularly for controlled single-object scenarios commonly encountered in automated sorting systems. In this work, we introduce $\textbf{SortScrews}$, a dataset for casewise visual classification of screws. The dataset contains 560 RGB images at $512\times512$ resolution covering six screw types and a background class. Images are captured using a standardized acquisition setup and include mild variations in lighting and camera perspective across four capture settings. To facilitate reproducible research and dataset expansion, we also provide a reusable data collection script that allows users to easily construct similar datasets for custom hardware components using inexpensive camera setups. We establish baseline results using transfer learning with EfficientNet-B0 and ResNet-18 classifiers pretrained on ImageNet. In addition, we conduct a well-explored failure analysis. Despite the limited dataset size, these lightweight models achieve strong classification accuracy, demonstrating that controlled acquisition conditions enable effective learning even with relatively small datasets. The dataset, collection pipeline, and baseline training code are publicly available at https://github.com/ATATC/SortScrews.

VideoTemp-o3: Harmonizing Temporal Grounding and Video Understanding in Agentic Thinking-with-Videos

✇cs.AI, q-bio.NC updates on arXiv.org
By: Wenqi Liu · Yunxiao Wang · Shijie Ma · Meng Liu · Qile Su · Tianke Zhang · Haonan Fan · Changyi Liu · Kaiyu Jiang · Jiankang Chen · Kaiyu Tang · Bin Wen · Fan Yang · Tingting Gao · Han Li · Yinwei Wei · Xuemeng Song
16 March 2026 at 12:00
arXiv:2602.07801v3 Announce Type: replace-cross Abstract: In long-video understanding, conventional uniform frame sampling often fails to capture key visual evidence, leading to degraded performance and increased hallucinations. To address this, recent agentic thinking-with-videos paradigms have emerged, adopting a localize-clip-answer pipeline in which the model actively identifies relevant video segments, performs dense sampling within those clips, and then produces answers. However, existing methods remain inefficient, suffer from weak localization, and adhere to rigid workflows. To solve these issues, we propose VideoTemp-o3, a unified agentic thinking-with-videos framework that jointly models video grounding and question answering. VideoTemp-o3 exhibits strong localization capability, supports on-demand clipping, and can refine inaccurate localizations. Specifically, in the supervised fine-tuning stage, we design a unified masking mechanism that encourages exploration while preventing noise. For reinforcement learning, we introduce dedicated rewards to mitigate reward hacking. Besides, from the data perspective, we develop an effective pipeline to construct high-quality long video grounded QA data, along with a corresponding benchmark for systematic evaluation across various video durations. Experimental results demonstrate that our method achieves remarkable performance on both long video understanding and grounding.

BitDance: Scaling Autoregressive Generative Models with Binary Tokens

✇cs.AI, q-bio.NC updates on arXiv.org
By: Yuang Ai · Jiaming Han · Shaobin Zhuang · Weijia Mao · Xuefeng Hu · Ziyan Yang · Zhenheng Yang · Yali Wang · Huaibo Huang · Xiangyu Yue · Hao Chen
16 March 2026 at 12:00
arXiv:2602.14041v2 Announce Type: replace-cross Abstract: We present BitDance, a scalable autoregressive (AR) image generator that predicts binary visual tokens instead of codebook indices. With high-entropy binary latents, BitDance lets each token represent up to $2^{256}$ states, yielding a compact yet highly expressive discrete representation. Sampling from such a huge token space is difficult with standard classification. To resolve this, BitDance uses a binary diffusion head: instead of predicting an index with softmax, it employs continuous-space diffusion to generate the binary tokens. Furthermore, we propose next-patch diffusion, a new decoding method that predicts multiple tokens in parallel with high accuracy, greatly speeding up inference. On ImageNet 256x256, BitDance achieves an FID of 1.24, the best among AR models. With next-patch diffusion, BitDance beats state-of-the-art parallel AR models that use 1.4B parameters, while using 5.4x fewer parameters (260M) and achieving 8.7x speedup. For text-to-image generation, BitDance trains on large-scale multimodal tokens and generates high-resolution, photorealistic images efficiently, showing strong performance and favorable scaling. When generating 1024x1024 images, BitDance achieves a speedup of over 30x compared to prior AR models. We release code and models to facilitate further research on AR foundation models. Code and models are available at: https://github.com/shallowdream204/BitDance.

LINC-AC092535.5 regulates MICAL2 mRNA level to inhibit p53-mediated ferroptosis in nasopharyngeal carcinoma

✇Oncogene - Issue - nature.com science feeds
By: Siyu Zhang · Wenyue Chen · Ji Yang · Zhen Ge · Yan Kong · Haiou Yan · Lili Bao · Ying Shan · Zhiyuan Yang · Wenhui Chen · Si Pan · Yiwen You · Bo You · Qicheng Zhang
14 March 2026 at 08:00

Oncogene, Published online: 14 March 2026; doi:10.1038/s41388-026-03714-y

LINC-AC092535.5 regulates MICAL2 mRNA level to inhibit p53-mediated ferroptosis in nasopharyngeal carcinoma
Received — 14 March 2026 ⏭

Multi-Omics Characterization of Lactate-Associated Molecular Subtypes in Lung Cancer Suggests a Role for DKK1 in Lactate-Linked Migration, Invasion, and Lactylation Programs

✇(Multiomics OR Omics) AND (Lung OR gastric OR Hepatocellular)
By: Hang Yu · Xiao-Bin An · Jin-Cheng Xu · Zhen Zhang · Long-Kai Yang · Long Qin · Qing-Sui Li · Chen-Hong Li · Xu Su · Dan Yang · Ning Wang · Jun-Nan Guo
14 March 2026 at 18:00

Cancers (Basel). 2026 Feb 25;18(5):735. doi: 10.3390/cancers18050735.

ABSTRACT

BACKGROUND: Lactate accumulation is increasingly recognized as a feature of tumor metabolic reprogramming that can coincide with immune dysregulation and aggressive phenotypes. The prognostic and immunologic relevance of lactate-associated heterogeneity in lung cancer remains to be clarified.

METHODS: We curated lactate-related genes and identified prognostic candidates in lung cancer cohorts. Consensus clustering was applied to define lactate-associated molecular subtypes, followed by characterization of survival and tumor microenvironment features. A LASSO-based gene signature was developed to generate an individual-level risk score and an integrated nomogram. Multi-omics analyses were used to evaluate concordance between transcriptomic and proteomic alterations. Single-cell transcriptomic data were analyzed to explore cellular heterogeneity in lactate-related programs. In vitro assays evaluated the response of candidate genes to lactate exposure and assessed cell migration and invasion under proliferation-inhibited conditions after genetic perturbation.

RESULTS: Two lactate-associated molecular subtypes were identified with distinct overall survival and divergent immune microenvironment features. Subtype 1 was associated with better outcomes and a more immune-inflamed profile, whereas Subtype 2 was associated with poorer outcomes and a myeloid-enriched, immunosuppressive contexture. Pathway analyses indicated subtype-associated differences in extracellular matrix-related processes and apoptosis-associated signaling. We developed an 11-gene prognostic signature and nomogram that stratified patients by risk across TCGA and GEO cohorts. Multi-omics integration highlighted ANLN, FGA, and DKK1 as consistently dysregulated at both transcript and protein levels. Among these candidates, DKK1 showed lactate-responsive induction in vitro. DKK1 perturbation altered lactate-enhanced migratory and invasive phenotypes and was accompanied by changes in intracellular lactate levels and global protein lactylation, supporting a potential feedforward relationship between lactate exposure, DKK1 expression, and lactylation.

CONCLUSIONS: This study characterizes lactate-associated molecular heterogeneity in lung cancer and provides a lactate-related subtype framework and prognostic risk model for patient stratification. The findings nominate DKK1 as a lactate-responsive candidate linked to migration/invasion phenotypes and lactate/lactylation changes in vitro.

PMID:41827671 | PMC:PMC12985219 | DOI:10.3390/cancers18050735

Multi-Omics Characterization of Lactate-Associated Molecular Subtypes in Lung Cancer Suggests a Role for DKK1 in Lactate-Linked Migration, Invasion, and Lactylation Programs

✇Omics In Lung
By: Hang Yu · Xiao-Bin An · Jin-Cheng Xu · Zhen Zhang · Long-Kai Yang · Long Qin · Qing-Sui Li · Chen-Hong Li · Xu Su · Dan Yang · Ning Wang · Jun-Nan Guo
14 March 2026 at 18:00

Cancers (Basel). 2026 Feb 25;18(5):735. doi: 10.3390/cancers18050735.

ABSTRACT

BACKGROUND: Lactate accumulation is increasingly recognized as a feature of tumor metabolic reprogramming that can coincide with immune dysregulation and aggressive phenotypes. The prognostic and immunologic relevance of lactate-associated heterogeneity in lung cancer remains to be clarified.

METHODS: We curated lactate-related genes and identified prognostic candidates in lung cancer cohorts. Consensus clustering was applied to define lactate-associated molecular subtypes, followed by characterization of survival and tumor microenvironment features. A LASSO-based gene signature was developed to generate an individual-level risk score and an integrated nomogram. Multi-omics analyses were used to evaluate concordance between transcriptomic and proteomic alterations. Single-cell transcriptomic data were analyzed to explore cellular heterogeneity in lactate-related programs. In vitro assays evaluated the response of candidate genes to lactate exposure and assessed cell migration and invasion under proliferation-inhibited conditions after genetic perturbation.

RESULTS: Two lactate-associated molecular subtypes were identified with distinct overall survival and divergent immune microenvironment features. Subtype 1 was associated with better outcomes and a more immune-inflamed profile, whereas Subtype 2 was associated with poorer outcomes and a myeloid-enriched, immunosuppressive contexture. Pathway analyses indicated subtype-associated differences in extracellular matrix-related processes and apoptosis-associated signaling. We developed an 11-gene prognostic signature and nomogram that stratified patients by risk across TCGA and GEO cohorts. Multi-omics integration highlighted ANLN, FGA, and DKK1 as consistently dysregulated at both transcript and protein levels. Among these candidates, DKK1 showed lactate-responsive induction in vitro. DKK1 perturbation altered lactate-enhanced migratory and invasive phenotypes and was accompanied by changes in intracellular lactate levels and global protein lactylation, supporting a potential feedforward relationship between lactate exposure, DKK1 expression, and lactylation.

CONCLUSIONS: This study characterizes lactate-associated molecular heterogeneity in lung cancer and provides a lactate-related subtype framework and prognostic risk model for patient stratification. The findings nominate DKK1 as a lactate-responsive candidate linked to migration/invasion phenotypes and lactate/lactylation changes in vitro.

PMID:41827671 | PMC:PMC12985219 | DOI:10.3390/cancers18050735

Inhibition of ZBTB7B-mediated ADPGK transcription by NEDD4 impedes glycolysis and progression of lung adenocarcinoma

✇Oncogenesis - nature.com science feeds
By: Fang-Peng Liu · Ting Huang · Hai-Yan Yang · Jian-Hua Zha · Chuan Xu · Cheng-Peng Qiu · Xing Wang · Zhuo Lu · Tao Wang
11 March 2026 at 08:00

Oncogenesis, Published online: 11 March 2026; doi:10.1038/s41389-026-00605-5

Inhibition of ZBTB7B-mediated ADPGK transcription by NEDD4 impedes glycolysis and progression of lung adenocarcinoma
Received — 11 March 2026 ⏭

Better Eyes, Better Thoughts: Why Vision Chain-of-Thought Fails in Medicine

✇cs.AI, q-bio.NC updates on arXiv.org
By: Yuan Wu · Zongxian Yang · Jiayu Qian · Songpan Gao · Guanxing Chen · Qiankun Li · Yu-An Huang · Zhi-An Huang
10 March 2026 at 12:00
arXiv:2603.06665v1 Announce Type: cross Abstract: Large vision-language models (VLMs) often benefit from chain-of-thought (CoT) prompting in general domains, yet its efficacy in medical vision-language tasks remains underexplored. We report a counter-intuitive trend: on medical visual question answering, CoT frequently underperforms direct answering (DirA) across general-purpose and medical-specific models. We attribute this to a \emph{medical perception bottleneck}: subtle, domain-specific cues can weaken visual grounding, and CoT may compound early perceptual uncertainty rather than correct it. To probe this hypothesis, we introduce two training-free, inference-time grounding interventions: (i) \emph{perception anchoring} via region-of-interest cues and (ii) \emph{description grounding} via high-quality textual guidance. Across multiple benchmarks and model families, these interventions improve accuracy, mitigate CoT degradation, and in several settings reverse the CoT--DirA inversion. Our findings suggest that reliable clinical VLMs require robust visual grounding and cross-modal alignment, beyond extending text-driven reasoning chains. Code is available \href{https://github.com/TianYin123/Better_Eyes_Better_Thoughts}{here}.

Revealing Behavioral Plasticity in Large Language Models: A Token-Conditional Perspective

✇cs.AI, q-bio.NC updates on arXiv.org
By: Liyuan Mao · Le Yu · Jing Zhou · Chujie Zheng · Bowen Yu · Chang Gao · Shixuan Liu · An Yang · Weinan Zhang · JunYang Lin
10 March 2026 at 12:00
arXiv:2603.08398v1 Announce Type: cross Abstract: In this work, we reveal that Large Language Models (LLMs) possess intrinsic behavioral plasticity-akin to chameleons adapting their coloration to environmental cues-that can be exposed through token-conditional generation and stabilized via reinforcement learning. Specifically, by conditioning generation on carefully selected token prefixes sampled from responses exhibiting desired behaviors, LLMs seamlessly adapt their behavioral modes at inference time (e.g., switching from step-by-step reasoning to direct answering) without retraining. Based on this insight, we propose Token-Conditioned Reinforcement Learning (ToCoRL), a principled framework that leverages RL to internalize this chameleon-like plasticity, transforming transient inference-time adaptations into stable and learnable behavioral patterns. ToCoRL guides exploration with token-conditional generation and keep enhancing exploitation, enabling emergence of appropriate behaviors. Extensive experiments show that ToCoRL enables precise behavioral control without capability degradation. Notably, we show that large reasoning models, while performing strongly on complex mathematics, can be effectively adapted to excel at factual question answering, which was a capability previously hindered by their step-by-step reasoning patterns.

Batch-of-Thought: Cross-Instance Learning for Enhanced LLM Reasoning

✇cs.AI, q-bio.NC updates on arXiv.org
By: Xuan Yang · Furong Jia · Roy Xie · Xiong Xi · Hengwei Bian · Jian Li · Monica Agrawal
10 March 2026 at 12:00
arXiv:2601.02950v2 Announce Type: replace Abstract: Current Large Language Model reasoning systems process queries independently, discarding valuable cross-instance signals such as shared reasoning patterns and consistency constraints. We introduce Batch-of-Thought (BoT), a training-free method that processes related queries jointly to enable cross-instance learning. By performing comparative analysis across batches, BoT identifies high-quality reasoning templates, detects errors through consistency checks, and amortizes computational costs. We instantiate BoT within a multi-agent reflection architecture (BoT-R), where a Reflector performs joint evaluation to unlock mutual information gain unavailable in isolated processing. Experiments across three model families and six benchmarks demonstrate that BoT-R consistently improves accuracy and confidence calibration while reducing inference costs by up to 61%. Our theoretical and experimental analysis reveals when and why batch-aware reasoning benefits LLM systems. Our code is available at https://github.com/xuanyang19/BoT

Multimodal Laryngoscopic Video Analysis for Assisted Diagnosis of Vocal Fold Paralysis

✇cs.AI, q-bio.NC updates on arXiv.org
By: Yucong Zhang · Xin Zou · Jinshan Yang · Wenjun Chen · Juan Liu · Faya Liang · Ming Li
10 March 2026 at 12:00
arXiv:2409.03597v4 Announce Type: replace-cross Abstract: This paper presents the Multimodal Laryngoscopic Video Analyzing System (MLVAS), a novel system that leverages both audio and video data to automatically extract key video segments and metrics from raw laryngeal videostroboscopic videos for assisted clinical assessment. The system integrates video-based glottis detection with an audio keyword spotting method to analyze both video and audio data, identifying patient vocalizations and refining video highlights to ensure optimal inspection of vocal fold movements. Beyond key video segment extraction from the raw laryngeal videos, MLVAS is able to generate effective audio and visual features for Vocal Fold Paralysis (VFP) detection. Pre-trained audio encoders are utilized to encode the patient voice to get the audio features. Visual features are generated by measuring the angle deviation of both the left and right vocal folds to the estimated glottal midline on the segmented glottis masks. To get better masks, we introduce a diffusion-based refinement that follows traditional U-Net segmentation to reduce false positives. We conducted several ablation studies to demonstrate the effectiveness of each module and modalities in the proposed MLVAS. The experimental results on a public segmentation dataset show the effectiveness of our proposed segmentation module. In addition, unilateral VFP classification results on a real-world clinic dataset demonstrate MLVAS's ability of providing reliable and objective metrics as well as visualization for assisted clinical diagnosis.

OTESGN: Optimal Transport-Enhanced Syntactic-Semantic Graph Networks for Aspect-Based Sentiment Analysis

✇cs.AI, q-bio.NC updates on arXiv.org
By: Xinfeng Liao · Xuanqi Chen · Lianxi Wang · Jiahuan Yang · Zhuowei Chen · Ziying Rong
10 March 2026 at 12:00
arXiv:2509.08612v3 Announce Type: replace-cross Abstract: Aspect-based sentiment analysis (ABSA) aims to identify aspect terms and determine their sentiment polarity. While dependency trees combined with contextual semantics provide structural cues, existing approaches often rely on dot-product similarity and fixed graphs, which limit their ability to capture nonlinear associations and adapt to noisy contexts. To address these limitations, we propose the Optimal Transport-Enhanced Syntactic-Semantic Graph Network (OTESGN), a model that jointly integrates structural and distributional signals. Specifically, a Syntactic Graph-Aware Attention module models global dependencies with syntax-guided masking, while a Semantic Optimal Transport Attention module formulates aspect-opinion association as a distribution matching problem solved via the Sinkhorn algorithm. An Adaptive Attention Fusion mechanism balances heterogeneous features, and contrastive regularization enhances robustness. Extensive experiments on three benchmark datasets (Rest14, Laptop14, and Twitter) demonstrate that OTESGN delivers state-of-the-art performance. Notably, it surpasses competitive baselines by up to +1.30 Macro-F1 on Laptop14 and +1.01 on Twitter. Ablation studies and visualization analyses further highlight OTESGN's ability to capture fine-grained sentiment associations and suppress noise from irrelevant context.
Received — 5 March 2026 ⏭

TTSR: Test-Time Self-Reflection for Continual Reasoning Improvement

✇cs.AI, q-bio.NC updates on arXiv.org
By: Haoyang He · Zihua Rong · Liangjie Zhao · Yunjia Zhao · Lan Yang · Honggang Zhang
5 March 2026 at 13:00
arXiv:2603.03297v1 Announce Type: cross Abstract: Test-time Training enables model adaptation using only test questions and offers a promising paradigm for improving the reasoning ability of large language models (LLMs). However, it faces two major challenges: test questions are often highly difficult, making self-generated pseudo-labels unreliable, and existing methods lack effective mechanisms to adapt to a model's specific reasoning weaknesses, leading to inefficient learning. To address these issues, we propose \textbf{TTSR}, a self-reflective test-time self-evolving training framework. TTSR employs a single pretrained language model that alternates between the roles of a \textit{Student} and a \textit{Teacher} at test time. The Student focuses on solving problems and learning from synthesized variant questions, while the Teacher analyzes the Student's failed reasoning trajectories, summarizes recurring reasoning weaknesses, and synthesizes targeted variant questions accordingly. This process guides the model to improve within a learnable regime through a continual self-evolving loop. Experimental results on multiple challenging mathematical reasoning benchmarks show that TTSR consistently improves reasoning performance and generalizes well across different model backbones and general-domain reasoning tasks. These findings suggest that teacher-mediated self-reflection provides an effective pathway for stable and continual reasoning improvement at test time.

Local Shapley: Model-Induced Locality and Optimal Reuse in Data Valuation

✇cs.AI, q-bio.NC updates on arXiv.org
By: Xuan Yang · Hsi-Wen Chen · Ming-Syan Chen · Jian Pei
5 March 2026 at 13:00
arXiv:2603.03672v1 Announce Type: cross Abstract: The Shapley value provides a principled foundation for data valuation, but exact computation is #P-hard due to the exponential coalition space. Existing accelerations remain global and ignore a structural property of modern predictors: for a given test instance, only a small subset of training points influences the prediction. We formalize this model-induced locality through support sets defined by the model's computational pathway (e.g., neighbors in KNN, leaves in trees, receptive fields in GNNs), showing that Shapley computation can be projected onto these supports without loss when locality is exact. This reframes Shapley evaluation as a structured data processing problem over overlapping support-induced subset families rather than exhaustive coalition enumeration. We prove that the intrinsic complexity of Local Shapley is governed by the number of distinct influential subsets, establishing an information-theoretic lower bound on retraining operations. Guided by this result, we propose LSMR (Local Shapley via Model Reuse), an optimal subset-centric algorithm that trains each influential subset exactly once via support mapping and pivot scheduling. For larger supports, we develop LSMR-A, a reuse-aware Monte Carlo estimator that remains unbiased with exponential concentration, with runtime determined by the number of distinct sampled subsets rather than total draws. Experiments across multiple model families demonstrate substantial retraining reductions and speedups while preserving high valuation fidelity.

Agent Data Protocol: Unifying Datasets for Diverse, Effective Fine-tuning of LLM Agents

✇cs.AI, q-bio.NC updates on arXiv.org
By: Yueqi Song · Ketan Ramaneti · Zaid Sheikh · Ziru Chen · Boyu Gou · Tianbao Xie · Yiheng Xu · Danyang Zhang · Apurva Gandhi · Fan Yang · Joseph Liu · Tianyue Ou · Zhihao Yuan · Frank Xu · Shuyan Zhou · Xingyao Wang · Xiang Yue · Tao Yu · Huan Sun · Yu Su · Graham Neubig
5 March 2026 at 13:00
arXiv:2510.24702v2 Announce Type: replace-cross Abstract: Public research results on large-scale supervised finetuning of AI agents remain relatively rare, since the collection of agent training data presents unique challenges. In this work, we argue that the bottleneck is not a lack of underlying data sources, but that a large variety of data is fragmented across heterogeneous formats, tools, and interfaces. To this end, we introduce the agent data protocol (ADP), a light-weight representation language that serves as an "interlingua" between agent datasets in diverse formats and unified agent training pipelines downstream. The design of ADP is expressive enough to capture a large variety of tasks, including API/tool use, browsing, coding, software engineering, and general agentic workflows, while remaining simple to parse and train on without engineering at a per-dataset level. In experiments, we unified a broad collection of 13 existing agent training datasets into ADP format, and converted the standardized ADP data into training-ready formats for multiple agent frameworks. We performed SFT on these data, and demonstrated an average performance gain of ~20% over corresponding base models, and delivers state-of-the-art or near-SOTA performance on standard coding, browsing, tool use, and research benchmarks, without domain-specific tuning. All code and data are released publicly, in the hope that ADP could help lower the barrier to standardized, scalable, and reproducible agent training.

Weight Space Representation Learning via Neural Field Adaptation

✇cs.AI, q-bio.NC updates on arXiv.org
By: Zhuoqian Yang, Mathieu Salzmann, Sabine S\"usstrunk
5 March 2026 at 13:00
arXiv:2512.01759v2 Announce Type: replace-cross Abstract: In this work, we investigate the potential of weights to serve as effective representations, focusing on neural fields. Our key insight is that constraining the optimization space through a pre-trained base model and low-rank adaptation (LoRA) can induce structure in weight space. Across reconstruction, generation, and analysis tasks on 2D and 3D data, we find that multiplicative LoRA weights achieve high representation quality while exhibiting distinctiveness and semantic structure. When used with latent diffusion models, multiplicative LoRA weights enable higher-quality generation than existing weight-space methods.

UniG2U-Bench: Do Unified Models Advance Multimodal Understanding?

✇cs.AI, q-bio.NC updates on arXiv.org
By: Zimo Wen · Boxiu Li · Wanbo Zhang · Junxiang Lei · Xiaoyu Chen · Yijia Fan · Qi Zhang · Yujiang Wang · Lili Qiu · Bo Li · Ziwei Liu · Caihua Shan · Yifan Yang · Yifei Shen
4 March 2026 at 13:00
arXiv:2603.03241v1 Announce Type: cross Abstract: Unified multimodal models have recently demonstrated strong generative capabilities, yet whether and when generation improves understanding remains unclear. Existing benchmarks lack a systematic exploration of the specific tasks where generation facilitates understanding. To this end, we introduce UniG2U-Bench, a comprehensive benchmark categorizing generation-to-understanding (G2U) evaluation into 7 regimes and 30 subtasks, requiring varying degrees of implicit or explicit visual transformations. Extensive evaluation of over 30 models reveals three core findings: 1) Unified models generally underperform their base Vision-Language Models (VLMs), and Generate-then-Answer (GtA) inference typically degrades performance relative to direct inference. 2) Consistent enhancements emerge in spatial intelligence, visual illusions, or multi-round reasoning subtasks, where enhanced spatial and shape perception, as well as multi-step intermediate image states, prove beneficial. 3) Tasks with similar reasoning structures and models sharing architectures exhibit correlated behaviors, suggesting that generation-understanding coupling induces class-consistent inductive biases over tasks, pretraining data, and model architectures. These findings highlight the necessity for more diverse training data and novel paradigms to fully unlock the potential of unified multimodal modeling.
❌