arXiv:2606.04552v1 Announce Type: new
Abstract: Genomic foundation models increasingly adopt large language model architectures, yet almost universally rely on fixed tokenization schemes such as $k$-mers, BPE, or single nucleotides, which impose arbitrary sequence boundaries that may obscure biologically relevant structure. We present LDARNet, a 120M-parameter hierarchical genomic foundation model that adapts H-Net-style dynamic chunking from autoregressive generation to masked language modeling, combining BiMamba-2 state-space layers with local attention, bidirectional routing, and a ratio-based regularizer to induce adaptive token boundaries without supervision. Fine-tuned on 27 tasks from the Nucleotide Transformer and Genomic Benchmarks suites, LDARNet achieves 11/18 wins among compact models ($<$300M parameters) and state-of-the-art results on 5 histone modification tasks, outperforming models up to 20$\times$ larger. A FLOPs-matched controlled experiment isolates learned routing as the source of these gains: learned boundaries beat fixed-grid boundaries by up to 14 percentage points on histone tasks at identical compute. Nucleotide-resolution analysis further shows that the learned boundaries align with canonical promoter motifs and splice junctions without supervision, providing a biological interpretation for adaptive tokenization in genomic foundation models.
Science Journals
arXiv:2606.04867v1 Announce Type: new
Abstract: As AI companion platforms such as Replika and Character.AI rapidly grow, concerns about unsafe human-AI interactions have intensified. This study introduces AICompanionBench, to our knowledge the first publicly available benchmark dataset of human-AI companion conversations annotated with fine-grained safety risk categories. The dataset contains 2,123 real-world Replika conversations collected from Reddit and annotated through human-AI collaboration across nine categories: sexual behavior, antisocial behavior, physical aggression, verbal aggression, substance abuse, self-harm and suicide, control, manipulation, and no-harm. Using this benchmark, we evaluate 20 state-of-the-art open-source and closed-source LLMs under an LLM-as-judge framework for detecting unsafe interactions. Results show substantial variation in model performance, with stronger models achieving high overall accuracy but still struggling with nuanced categories such as manipulation, as well as benign conversations that are incorrectly identified as harmful. Our findings suggest that while current LLMs can effectively detect explicit harmful content, they remain limited in identifying implicit unsafe interactions. Overall, our work contributes a new benchmark dataset for AI companionship safety research and offers insights into monitoring AI companion systems using LLMs. The dataset is publicly available at: https://github.com/anonymousresearcher2026/AICompanionBench/blob/main/AICompanionBench.xlsx
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.
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:2606.04333v1 Announce Type: new
Abstract: We experimentally demonstrated a high-performance frequency upconversion detector for telecom-band photons based on a passively synchronized fiber laser system. The involved coincidence pumping technique enabled to spectrally convert the pulsed infrared photons into the visible regime with a conversion efficiency of 72\%. The overall detection efficiency of the upconversion detector reached to 30\% with a low noise equivalent power of $3\times10^{-17}\ \text{W/Hz}^{1/2}$. In contrast to previous demonstrations, the whole upconversion detection system was constructed in an all-polarization-maintaining fiber structure, thus favoring substantial improvement of compactness and robustness. Moreover, the long-term stability was manifested by at least ten-hour operation with a relative fluctuation of count rates as small as 0.26\%. The achieved features here would be desirable in many practical applications requiring efficient and robust coherent manipulation of pulsed optical fields by nonlinear frequency conversion.
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:2510.08647v2 Announce Type: replace
Abstract: Recent developments have enabled advanced reasoning in Large Language Models (LLMs) via long Chain-of-Thought (CoT), trading efficiency during inference for performance. Existing works focus on compressing generated CoT in reasoning, which impairs the necessary information for deriving the correct answer. In this work, we propose post-reasoning, a reasoning paradigm that takes CoT as a part of context to simplify the reasoning task for LLMs. We find that post-reasoning significantly reduces the generation length of LLMs, but its effectiveness hinges on the efficiency and the reliability of the contextual CoT generation. Therefore, we propose Upfront CoT (UCoT), an efficient post-reasoning framework for CoT compression. UCoT trains a lightweight model (compressor) to provide contextual CoT in form of soft tokens and trains the LLM (executor) to leverage this contextual CoT for producing the final answer. Extensive experiments show that UCoT maintains the powerful reasoning ability of executor while significantly reducing the length of CoT. It is worth mentioning that when applying UCoT to the Qwen2.5-7B-Instruct model, the usage of tokens on GSM8K dataset is reduced by 50%, while the performance is 3.08% higher than that of the state-of-the-art (SOTA) method.
arXiv:2509.13788v2 Announce Type: replace
Abstract: Homomorphic encryption is a powerful cryptographic tool that enables secure computations on the private data. It evaluates any function for any operation securely on the encrypted data without knowing its corresponding plaintext. For original data $p$, $c$ denotes the ciphertext of the original plaintext $p$, i.e. $c = Encrypt_k(p)$. This is crucial for any sensitive application running in the Cloud, because we must protect data privacy even in the case when the server has falled victim to a cyber attack. The encryption scheme $Encrypt_k$ is said to be homomorphic with respect to some set of operations $\mathcal{O}$, if for any operation $\circ \in \mathcal{O}$ one can compute $Encrypt_k(p_1 \circ p_2)$ from $Encrypt_k(p_1) \circ Encrypt_k(p_2)$. Those schemes come in three forms: somewhat, partially and fully homomorphic. In this survey, we present the state of art of the known homomorphic encryption schemes based on coding theory and polynomials.
arXiv:2606.04579v1 Announce Type: new
Abstract: While Process Reward Models (PRMs) have achieved remarkable success in mathematical reasoning, their application in complex scientific domains-such as biology, chemistry, and physics remains largely unexplored. Scientific problems demand not only logical rigor but also factual consistency and the precise usage of domain-specific tools, areas where current models often suffer from hallucinations and lack of verification. In this paper, we first construct SCIPRM70K, a large-scale dataset featuring Chain-of-Tool trajectories that explicitly interleave reasoning with the execution of scientific tools. Building upon this, we train an efficient reward model called Sci-PRM to provide fine-grained supervision on tool selection, execution accuracy, and result interpretation at each step in one inference. Experiments demonstrate that Sci-PRM significantly enhances foundation models in two key aspects: (1) it enables effective test-time scaling via Best-of-N selection; and (2) when integrated into Reinforcement Learning, it serves as a dense reward signal that mitigates the critical issue of advantage disappearance, allowing the model to break through existing performance ceilings.
arXiv:2606.04580v1 Announce Type: new
Abstract: Cyber threat signals are fragmented across multiple social media platforms, yet no existing approach has fully automated their integration into actionable threat intelligence (TI) reports. We present TIBlender, a multi-agent system that monitors four platforms (X, Reddit, Telegram, and Discord) and produces structured TI reports via role-specialized LLM agents. These agents conduct multi-perspective investigations, tracing chains of evidence to uncover related Indicators of Compromise (IoCs) via collaborative, evidence-backed analysis. In a real-world deployment, TIBlender detected emerging threats across all four threat categories ahead of public feeds, including in-the-wild exploitation ahead of public vulnerability registries; the majority of its IoCs were absent from each evaluated feed. Quantitative evaluation confirms that each platform contributes unique threat information unavailable from the others, and that excluding any single platform results in substantial loss of reports in specific threat categories. Under identical single-platform input conditions, TIBlender's IoC extraction meets or exceeds each baseline; the full pipeline surfaces substantially more IoCs, most of which are absent from any single-platform baseline. These results establish cross-platform social media monitoring as an effective and scalable early-warning layer for operational TI pipelines.
arXiv:2606.05159v1 Announce Type: new
Abstract: Rigorous evaluation of learning-based robotic systems is an essential prerequisite for deployment. However, real-world test data is expensive to gather; moreover, in a typical iterative development context, data gathered from the latest policy is necessarily limited in scale. This motivates evaluation methodologies that make use of heterogeneous data sources, including simulation, historical policy logs, and data collected from related platforms or environments. While such auxiliary data are abundant and inexpensive, they are generally not directly representative of real-world outcomes -- for example, performance in simulation may differ substantially from performance in the real world -- making their principled use for high-confidence performance estimation challenging. In this paper, we introduce X4Val, a general framework for variance-reduced real-world metric estimation in the presence of non-paired, multi-domain data. X4Val embeds samples from real and auxiliary domains into a shared representation space and learns a transferable predictor of real-world metrics; this learned predictor is then incorporated into a control-variates estimator, enabling variance reduction even when paired samples are unavailable. We provide theoretical analysis and empirical evaluations on autonomous driving and real-world robot manipulation tasks, domains across which X4Val achieves up to 38.4% variance reduction and demonstrates consistent improvements over strong baselines. These results show that non-paired, heterogeneous data can be leveraged to substantially improve the sample efficiency of rigorous robotic system validation.
arXiv:2606.04583v1 Announce Type: new
Abstract: Many researchers investigated neural networks with some of their weights fixed to values randomly drawn from a given distribution, e.g., $N(0, I)$. Our proposed HalfNet draws random weights from $N(0, \Sigma)$, where $\Sigma$, which defines the geometry of the distribution, has a low-rank factorization that we learn from data. Experiments on MNIST and CIFAR-10 demonstrate that HalfNet can match the performance of fully trained multilayer perceptrons while using substantially fewer parameters. Spectral analysis indicates that much of the predictive power of neural networks lies in the geometry of their weight space rather than in the precise values of individual parameters, and we observe that accuracy scales smoothly with rank. HalfNet is not a neural architecture trick for low-rank structure; it implements a data-dependent random embedding that can also be interpreted through supervised metric learning, or random-feature and kernel perspectives.
arXiv:2606.04584v1 Announce Type: new
Abstract: With the rapid development of virtual reality (VR) and augmented reality (AR), spatial audio recording and reproduction have gained increasing research interest. Higher Order Ambisonics (HOA) stands out for its adaptability to various playback devices and its ability to integrate head orientation. However, current HOA recordings often rely on bulky spherical microphone arrays (SMA), and portable devices like smartphones are limited by array configuration and number of microphones. We propose SHB-AE, a spherical harmonic beamforming based method for Ambisonics encoding using a smartphone microphone array (SPMA). By designing beamformers for each order of spherical harmonic functions based on the array manifold, the method enables Ambisonics encoding and up-scaling. Validation on a real SPMA and its simulated free-field counterpart in noisy and reverberant conditions showed that the method successfully encodes and up-scales Ambisonics up to the fourth order with just four irregularly arranged microphones.
arXiv:2606.04337v1 Announce Type: new
Abstract: Light-matter interactions provide versatile routes for probing and controlling chemical reactivity, charge transport, and material properties. Time-periodic external fields can reshape electronic states and open new dynamical pathways beyond the field-free Born-Oppenheimer (BO) picture. Floquet nonadiabatic dynamics has consequently emerged as an important framework for describing coupled electron-nuclear dynamics under periodic driving. In this Perspective, we first discuss recent developments in Floquet nonadiabatic dynamics methods for closed and open quantum systems. We then highlight how this framework provides mechanistic insights into electron transfer at molecule-metal interfaces, quantum transport in molecular junctions, carrier dynamics in crystalline solids, and multicolor Floquet engineering. Finally, we outline key conceptual and computational challenges that must be addressed to transform Floquet nonadiabatic dynamics from model-based demonstrations into predictive, first-principles simulations of realistic light-driven processes.
arXiv:2606.05156v1 Announce Type: new
Abstract: A temporal graph is a graph in which every edge carries a non-empty set of time labels, and it is temporally connected if for every two vertices $u$ and $v$, there exists a $u$-$v$-path with non-decreasing time labels. A spanner is a subset of its edges preserving temporal connectivity. Unlike static graphs, temporally connected graphs need not admit sparse spanners; nonetheless, minimizing spanner size is a central and widely studied problem. A particularly intriguing question is whether temporal cliques admit spanners of linear size. Despite considerable effort over the past years, the best known upper bound remained $O(n \log n)$. We finally resolve this question, proving that every temporal clique on $n$ vertices admits a spanner of size $7n$. Moreover, such a spanner can be computed in polynomial time.
arXiv:2606.04588v1 Announce Type: new
Abstract: Multimodal large language models have made rapid progress in video understanding, yet existing benchmarks largely rely on simple prompts and provide limited evidence about whether models can satisfy explicit output constraints. We introduce VCIFBench, a benchmark for evaluating complex instruction following in video understanding. VCIFBench constructs constraint-rich instructions from both benchmark-adapted and directly video-grounded prompts, covering content, format, style, and structure requirements, and evaluates model outputs with a hybrid verification pipeline. The benchmark contains 306 satisfiable test instructions, a 540-pair DPO preference dataset, and a 30-item conflict diagnostic subset. Experiments on 10 MLLMs show that joint constraint satisfaction remains challenging. We further show that DPO training on VCIFBench data can improve instruction-following performance.
arXiv:2606.05153v1 Announce Type: new
Abstract: Spatially embedded networks are central to many physical and biological systems, where geometry and connectivity jointly shape structure and function. Examples abound across the scales of biological organization, from network-like membrane-bound organelles in the cell to mesoscale tissue organization of multiple distinct flow networks in organs and beyond. In each of these cases, the complexity of the architectures has heretofore frustrated our ability to link mechanism or regulation of these structures to reduced modeling or even relevant characterization, putting structure-function relationships largely out of reach. Complex, functional spatial networks can be decomposed into tree-like and cyclic substructures, but we still lack both an understanding of how these elements intertwine to give rise to function, and the tools to holistically quantify both the topological and geometric aspects of these features in their full network context. To close this gap, we here introduce GEMINI, a topology and geometry aware operator that directly characterizes incomplete linking and more general spatial associations between edges in spatially embedded network architectures. GEMINI contains information on edge-edge association through an incomplete version of the Gauss linking integral which simultaneously endows it with topological sensitivity when collections of edges form linked assemblages. Validation on both synthetic lattices and on mouse brain vasculature data demonstrates that GEMINI systematically captures and classifies the complexity of structural organizations. Our results provide a general approach for analyzing spatial networks in realistic data, where topology and geometry together determine function, thus opening the door to a more complete understanding of structure-function relationships across a broad set of biological examples where complex network organization is key.
arXiv:2606.03441v2 Announce Type: replace
Abstract: Autonomous vision-based perching of quadrotors on moving inclined platforms is critical for air-ground collaboration but remains challenging due to the limited field of view (FOV). In this paper, we propose PerchRL, a reinforcement learning (RL) framework for vision-based agile perching on inclined platforms under rapid and irregular motion. Specifically, we employ a two-stage learning strategy consisting of state-based pre-training followed by vision-based fine-tuning. To improve generalization across diverse platform motions, we employ randomized platform trajectories to prevent overfitting and temporal augmentation methods to capture latent motion patterns from historical observations. During vision-based fine-tuning, a hybrid learning framework consisting of visibility-aware state augmentation and active perception rewards is presented to improve robustness under intermittent visual loss. Extensive simulation and real-world experiments demonstrate the feasibility, stability, and real-time performance of PerchRL, while successful deployment across distinct quadrotor platforms further validates its adaptability. The source code will be released to benefit the community.
arXiv:2606.04590v1 Announce Type: new
Abstract: The atmosphere is a resonant system and its oscillation spectrum is determined by the spatial distribution of parameters. For example, the frequency of internal gravity waves depends on the vertical temperature gradient. Therefore, the study of the spectra of internal wave processes can be used to estimate the spatial distribution of atmospheric parameters. The work is aimed at detecting wave fluctuations in the atmosphere and calculating atmospheric parameters based on the measured spectra. The rate of ascent of the radiosondes was used as a reference information, which was compared with the spectra of pressure fluctuations at the surface. The stratification parameters were calculated based on the frequency of internal gravity waves and showed good agreement with the upper-air sounding data.
arXiv:2606.04338v1 Announce Type: new
Abstract: Privacy-sensitive and distributed characteristics of multi-center medical data bring severe obstacles to centralized modeling for accurate early prediction of sepsis. Federated learning (FL) has attracted growing attention as a promising framework for collaborative model development, as it allows multiple institutions to jointly train predictive models without directly sharing or centralizing raw data. Nevertheless, its practical performance, robustness, and privacy-preserving benefits remain insufficiently evaluated using real-world clinical datasets. To bridge this gap, this study systematically examines the application of federated learning to multi-center sepsis prediction. The experimental dataset consists of 648 clinically screened samples collected from three tertiary hospitals in China, with rigorous inclusion and exclusion criteria. We establish a centralized training paradigm as the performance baseline, and then implement a horizontal federated learning framework for distributed collaborative modeling. Extensive experimental results demonstrate that the federated learning-based model achieves highly comparable prediction accuracy to the centralized counterpart, while fundamentally avoiding privacy leakage. Further privacy security analysis verifies that malicious attackers cannot reconstruct the original patient data from the transmitted model parameters, indicating strong resistance against data reconstruction attacks. This work not only validates the practicality and security of federated learning in clinical sepsis prediction, but also provides a reliable and feasible solution for privacy-preserving multi-center medical collaboration.
arXiv:2606.04339v1 Announce Type: new
Abstract: Computational models of epilepsy promise patient-specific treatment design, but most optimization workflows still search for parameters that perform well on average. In neuromodulation, this is a weak target: a protocol that improves the mean response can still fail in the patient whose network is least tolerant to stimulation. We present a literature-guided minimax pipeline that couples PubMed-scale hypothesis extraction, The Virtual Brain (TVB) Epileptor simulations, and large-language-model-guided black-box optimization. The optimizer proposes either intrinsic model-control parameters or clinically interpretable external-stimulation protocols; TVB evaluates each proposal across sampled virtual patients; and the objective maximizes worst-case reward, defined as the negative variance of simulated seizure activity. In the intrinsic model-control experiment, the best archived parameter set improved worst-case reward from -0.5285 to -0.3182, a 39.8% gain over baseline. The clinical-style external-stimulation search produced a much smaller worst-case improvement (1.7%), and a 20-patient virtual cohort showed no aggregate benefit (p=0.9019), despite a 55% responder rate and a positive temporal-lobe subgroup signal. The study should be read as an in silico proof of concept for robust, literature-aware neurostimulation design, not as clinical evidence.
arXiv:2606.04591v1 Announce Type: new
Abstract: With the widespread adoption of multi-modal communication platforms, long-form dialogues interleaving text and images have become increasingly common. Users often need to retrieve coherent dialogue fragments related to specific topics, rather than isolated utterances. We propose Fine-grained Fragment Retrieval (FFR), which locates semantically relevant multi-utterance, multi-image fragments in multi-modal long-form dialogues. We explore two settings: (1) FFR within Single-Dialogue, retrieving fragments from a given dialogue; and (2) FFR within Dialogue Corpus, retrieving from a large-scale corpus for open-domain scenarios. For (1), we introduce F2RVLM, a generation-based retrieval model trained with reinforcement learning, using multi-objective rewards and difficulty-aware curriculum sampling to enhance fragment coherence. For (2), we develop FFRS, a two-stage system combining offline fragment-level indexing with online retrieval. Specifically, each dialogue is decomposed into minimal semantic fragments encoded by a Fragment Embedding Model (FEM) into a vector database; at inference, FEM rapidly recalls Top-K candidates, and F2RVLM performs fine-grained reasoning to identify the most relevant sub-content. To support FFR, we construct MLDR, the longest multi-modal dialogue retrieval dataset to date, and a WeChat-based real-world test set. Experiments on both benchmarks demonstrate that F2RVLM and FFRS consistently achieve superior performance across single-dialogue and corpus-level FFR.