Daily AI Paper Discovery · Agent / RAG / Knowledge Graph https://alloevil.github.io/AI-Paper-Daily zh-CN Sun, 16 Aug 2026 04:41:59 +0000 The expanding operational capabilities of large language model (LLM) agents introduce sophisticated security threats. Runtime defenses have emerged as an effective approach to mitigating these risks b... http://arxiv.org/abs/2608.12977v1 http://arxiv.org/abs/2608.12977v1#2026-08-16 Sun, 16 Aug 2026 18:00:00 +0000 The performance of large language model (LLM)-based multi-agent systems (MAS) largely depends on effective communication topologies. Existing topology generation methods, however, typically learn comm... http://arxiv.org/abs/2608.12921v1 http://arxiv.org/abs/2608.12921v1#2026-08-16 Sun, 16 Aug 2026 18:00:00 +0000 Agent-memory systems increasingly buy retrieval quality with structure, transforming raw conversation histories into summaries, embeddings, trees, or knowledge graphs before any question is asked. We ... http://arxiv.org/abs/2608.12888v1 http://arxiv.org/abs/2608.12888v1#2026-08-16 Sun, 16 Aug 2026 18:00:00 +0000 Self-improving LLM agents convert successful trajectories into persistent cross-task state. An unsafe success can thereby become reusable policy after its triggering input disappears. Skill evolution ... http://arxiv.org/abs/2608.12851v1 http://arxiv.org/abs/2608.12851v1#2026-08-16 Sun, 16 Aug 2026 18:00:00 +0000 Analog circuit design is a time-consuming, iterative process in a nonlinear and high-dimensional design space that relies heavily on expert intuition. Among recent developments, LLMs have introduced a... http://arxiv.org/abs/2608.13472v1 http://arxiv.org/abs/2608.13472v1#2026-08-15 Sat, 15 Aug 2026 18:00:00 +0000 Large language model based multi-agent systems usually communicate in text, i.e., using discrete tokens. However, text introduces a discrete bottleneck. Converting the sender's continuous hidden state... http://arxiv.org/abs/2608.13317v1 http://arxiv.org/abs/2608.13317v1#2026-08-15 Sat, 15 Aug 2026 18:00:00 +0000 As biomedical research increasingly relies on data-intensive tools, the quality and utility of datasets are critical. Challenges such as imbalances, biases, and ethical or legal constraints often limi... http://arxiv.org/abs/2608.13256v1 http://arxiv.org/abs/2608.13256v1#2026-08-15 Sat, 15 Aug 2026 18:00:00 +0000 Reinforcement learning with verifiable rewards (RLVR) offers a verifier-bounded performance ceiling for training multi-turn tool-use agents, yet its trajectory-level credit assignment conflates hetero... http://arxiv.org/abs/2608.13179v1 http://arxiv.org/abs/2608.13179v1#2026-08-15 Sat, 15 Aug 2026 18:00:00 +0000 Agent skills are crucial external instructions that enable language agents to execute long procedural tasks such as coding or document processing. Existing agent skills are primarily created through h... http://arxiv.org/abs/2608.13173v1 http://arxiv.org/abs/2608.13173v1#2026-08-15 Sat, 15 Aug 2026 18:00:00 +0000 When a security researcher publishes a report on a cyberattack, detection engineers are supposed to turn it into working detection rules. In practice, most automated attempts at this only extract the ... http://arxiv.org/abs/2608.13050v1 http://arxiv.org/abs/2608.13050v1#2026-08-15 Sat, 15 Aug 2026 18:00:00 +0000 Organizational decisions are co-created while evidence, constraints, and human priorities continue to evolve. In conventional transcript-based multi-agent systems, humans typically provide an initial ... http://arxiv.org/abs/2608.13046v1 http://arxiv.org/abs/2608.13046v1#2026-08-15 Sat, 15 Aug 2026 18:00:00 +0000 The field of bioinformatics struggles with legacy code - old code that is commonly used but may no longer have a maintainer, or may be written in an now-unfamiliar language (e.g. Perl, Fortran). This ... http://arxiv.org/abs/2608.13029v1 http://arxiv.org/abs/2608.13029v1#2026-08-15 Sat, 15 Aug 2026 18:00:00 +0000 Knowledge-base question answering (KBQA) systems rely on effective retrieval and reasoning mechanisms to generate accurate answers from external knowledge sources. However, developing reliable KBQA sy... http://arxiv.org/abs/2608.13004v1 http://arxiv.org/abs/2608.13004v1#2026-08-15 Sat, 15 Aug 2026 18:00:00 +0000 Long-horizon LLM agents must preserve information from past interactions to support future tasks. Existing memory systems typically rely on eager consolidation, invoking LLMs after each interaction to... http://arxiv.org/abs/2608.12990v1 http://arxiv.org/abs/2608.12990v1#2026-08-15 Sat, 15 Aug 2026 18:00:00 +0000 We present DreamX-Phi 1.0, an action-conditioned video world model for robotic manipulation that, given an observed frame, a language instruction, and a prescribed action sequence comprising end-effec... https://arxiv.org/abs/2608.13489 https://arxiv.org/abs/2608.13489#2026-08-14 Fri, 14 Aug 2026 18:00:00 +0000 Scientific discovery increasingly requires AI systems that can reason over scientific evidence of heterogeneous modalities, interact with scientific tools and environments, and sustain progress across... https://arxiv.org/abs/2608.13505 https://arxiv.org/abs/2608.13505#2026-08-14 Fri, 14 Aug 2026 18:00:00 +0000 Spatial intelligence is becoming a foundation for embodied agents, robotic planning, and multimodal assistants. To improve the spatial reasoning ability of VLM agents, existing work has mainly followe... https://arxiv.org/abs/2608.12743 https://arxiv.org/abs/2608.12743#2026-08-14 Fri, 14 Aug 2026 18:00:00 +0000 Transforming multimodal sources into condensed and structured media outputs can be fundamentally conceptualized as a long-horizon agentic process centered on a model-harness system. While an ideal har... https://arxiv.org/abs/2608.13560 https://arxiv.org/abs/2608.13560#2026-08-14 Fri, 14 Aug 2026 18:00:00 +0000 Video world models simulate future states conditioned on current observations and user actions. Recent systems have demonstrated impressive video consistency and action controllability over long seque... https://arxiv.org/abs/2608.13552 https://arxiv.org/abs/2608.13552#2026-08-14 Fri, 14 Aug 2026 18:00:00 +0000 Compositional reliability bounds for multi-agent systems multiply component reliabilities, a step licensed by a conditional-independence assumption that is routinely stated and rarely tested. We test ... http://arxiv.org/abs/2608.12895v1 http://arxiv.org/abs/2608.12895v1#2026-08-14 Fri, 14 Aug 2026 18:00:00 +0000 Interactive world models must support persistent memory, responsive interaction, and long-horizon generation, yet these requirements place conflicting demands on the model. Maintaining history in the ... https://arxiv.org/abs/2608.13546 https://arxiv.org/abs/2608.13546#2026-08-14 Fri, 14 Aug 2026 18:00:00 +0000 Large-scale manipulation data is essential for robot learning, yet collecting robot demonstrations remains expensive and difficult to scale. Meanwhile, abundant egocentric human manipulation videos pr... https://arxiv.org/abs/2608.13049 https://arxiv.org/abs/2608.13049#2026-08-14 Fri, 14 Aug 2026 18:00:00 +0000 Recent advances in foundation models have enabled AI scientists to automate increasingly complete research workflows, from hypothesis generation and code execution to manuscript preparation. Yet workf... https://arxiv.org/abs/2608.13558 https://arxiv.org/abs/2608.13558#2026-08-14 Fri, 14 Aug 2026 18:00:00 +0000 Talking-video character replacement requires coordinated transfer of appearance and voice while preserving the source motion, scene, linguistic content, and audio-video timing. Existing methods use se... https://arxiv.org/abs/2608.11752 https://arxiv.org/abs/2608.11752#2026-08-14 Fri, 14 Aug 2026 18:00:00 +0000 Recent work on distillation transfers the capabilities of large models to smaller ones often by updating the latter's parameters, through teacher forcing, on-policy distillation, and related training-... https://arxiv.org/abs/2608.12307 https://arxiv.org/abs/2608.12307#2026-08-13 Thu, 13 Aug 2026 18:00:00 +0000 Turning a research idea into a complete paper requires more than text generation: the system must retrieve literature, design and execute experiments, revise claims according to evidence, produce publ... https://arxiv.org/abs/2608.11924 https://arxiv.org/abs/2608.11924#2026-08-13 Thu, 13 Aug 2026 18:00:00 +0000 Previsualization is an intermediate layer between ideas and production in film, games, architecture, and urban design. It lets creators iteratively refine scenes, actions, cameras, and spatial-tempora... https://arxiv.org/abs/2608.12314 https://arxiv.org/abs/2608.12314#2026-08-13 Thu, 13 Aug 2026 18:00:00 +0000 LLM-agent services repeatedly execute small deterministic transitions between model and tool calls: route an outcome, update state, and emit the next effect. We ask when this control path exposes enou... http://arxiv.org/abs/2608.12123v1 http://arxiv.org/abs/2608.12123v1#2026-08-13 Thu, 13 Aug 2026 18:00:00 +0000 The Seungjeongwon Ilgi, a UNESCO Memory of the World record, is only 37.4% translated, and the most conspicuous failure mode in automatic translation is the person name -- a misread name corrupts the ... http://arxiv.org/abs/2608.11843v1 http://arxiv.org/abs/2608.11843v1#2026-08-13 Thu, 13 Aug 2026 18:00:00 +0000 Videos captured through glass often contain reflections that degrade visual quality and interfere with downstream vision tasks. Although single-image reflection removal has been extensively studied, v... https://arxiv.org/abs/2608.11562 https://arxiv.org/abs/2608.11562#2026-08-13 Thu, 13 Aug 2026 18:00:00 +0000 Agentic systems powered by large language models (LLMs) have opened new opportunities for business ideation. Yet existing approaches remain confined to a text-only paradigm, despite the inherently mul... https://arxiv.org/abs/2608.11616 https://arxiv.org/abs/2608.11616#2026-08-13 Thu, 13 Aug 2026 18:00:00 +0000 Complex software systems develop over timescales that exceed the lifespan of any individual coding agent. Most agentic software systems preserve continuity through persistent sessions, memories, manag... https://arxiv.org/abs/2608.10450 https://arxiv.org/abs/2608.10450#2026-08-13 Thu, 13 Aug 2026 18:00:00 +0000 Creative agents still lack an effective way to learn from high-quality human films, limiting their ability to produce cinematic-grade videos. A key challenge is the absence of a structured video repre... http://arxiv.org/abs/2608.12313v1 http://arxiv.org/abs/2608.12313v1#2026-08-13 Thu, 13 Aug 2026 18:00:00 +0000 Dynamic Master Logic (DML) provides a hierarchical framework for representing system behavior by linking functional objectives to underlying structural elements. However, DML construction typically re... http://arxiv.org/abs/2608.12304v1 http://arxiv.org/abs/2608.12304v1#2026-08-13 Thu, 13 Aug 2026 18:00:00 +0000 Does the cooperative bias documented for Western frontier LLM agents extend to a different alignment lineage, and should the Chinese models that embody it be treated as a single bloc or as distinct la... http://arxiv.org/abs/2608.10262v1 http://arxiv.org/abs/2608.10262v1#2026-08-12 Wed, 12 Aug 2026 18:00:00 +0000 Enterprises are deploying autonomous AI agents faster than they can govern them, and prevailing approaches stretch a single discipline, typically DevSecOps built for deterministic automation, across e... http://arxiv.org/abs/2608.10153v1 http://arxiv.org/abs/2608.10153v1#2026-08-12 Wed, 12 Aug 2026 18:00:00 +0000 Omni-modal dialogue models can understand multimodal inputs and synthesize spoken replies, yet their responses remain visually disembodied. We introduce Ex-Omni-2D, an omni-modal dialogue framework th... https://arxiv.org/abs/2608.10720 https://arxiv.org/abs/2608.10720#2026-08-12 Wed, 12 Aug 2026 18:00:00 +0000 Fréchet distance has recently emerged as an effective distribution-level objective for generator post-training, complementing the conventional sample-level diffusion and flow-matching losses. However,... https://arxiv.org/abs/2608.11205 https://arxiv.org/abs/2608.11205#2026-08-12 Wed, 12 Aug 2026 18:00:00 +0000 Self-evolving agents accumulate reusable skills by appending successful procedures and failure fixes. Over time, the same requirement is often restated in several branches, examples, and warnings, whi... https://arxiv.org/abs/2608.11079 https://arxiv.org/abs/2608.11079#2026-08-12 Wed, 12 Aug 2026 18:00:00 +0000 Large language models (LLMs) are increasingly deployed as mobile assistants, where a key challenge is leveraging personal information scattered across multiple applications (apps) to complete user ins... https://arxiv.org/abs/2608.10692 https://arxiv.org/abs/2608.10692#2026-08-12 Wed, 12 Aug 2026 18:00:00 +0000 We study reference-free post-training for multilingual machine translation with open large language models. Starting from the supervised-finetuned MiLMMT-46-v0.1 models, we apply Group Relative Policy... https://arxiv.org/abs/2608.10812 https://arxiv.org/abs/2608.10812#2026-08-12 Wed, 12 Aug 2026 18:00:00 +0000 Visual document retrieval (VDR) is dominated by multi-billion-parameter models that are slow to index at full corpus scale and expensive to serve. Prior compression routes either train a smaller multi... https://arxiv.org/abs/2608.10636 https://arxiv.org/abs/2608.10636#2026-08-12 Wed, 12 Aug 2026 18:00:00 +0000 4D generation synthesizes dynamic 3D scenes from conditions such as text or images. Existing methods either reconstruct generated RGB videos with a separate 4D model or adapt a particular video genera... https://arxiv.org/abs/2608.10744 https://arxiv.org/abs/2608.10744#2026-08-12 Wed, 12 Aug 2026 18:00:00 +0000 Real-world data science involves long-horizon workflows that span data wrangling, exploration, modeling, visualization, and validation, and require coordinated use of tools such as notebooks, IDEs, te... https://arxiv.org/abs/2608.10366 https://arxiv.org/abs/2608.10366#2026-08-12 Wed, 12 Aug 2026 18:00:00 +0000 As AI coding agents take on increasingly complex, long-horizon software engineering tasks, existing benchmarks are rapidly saturating and their evaluation quality has come under serious scrutiny: a re... https://arxiv.org/abs/2608.09802 https://arxiv.org/abs/2608.09802#2026-08-11 Tue, 11 Aug 2026 18:00:00 +0000 Macaron-V1 is an open agent-model family for experiential intelligence: learning from experience in real environments and continuing to learn after deployment. It is organized around two system goals.... https://arxiv.org/abs/2608.09819 https://arxiv.org/abs/2608.09819#2026-08-11 Tue, 11 Aug 2026 18:00:00 +0000 We introduce Sci-VBench, a comprehensive benchmark for evaluating knowledge- and reasoning-intensive video generation across scientific domains. It contains 1,253 expert-annotated examples spanning 60... https://arxiv.org/abs/2608.09873 https://arxiv.org/abs/2608.09873#2026-08-11 Tue, 11 Aug 2026 18:00:00 +0000 We introduce Motif 3, a decoder-only Mixture-of-Experts language model with 314 billion total parameters and 13.2 billion activated per token. Each sparse MoE layer contains 384 routed experts, with e... https://arxiv.org/abs/2608.09119 https://arxiv.org/abs/2608.09119#2026-08-11 Tue, 11 Aug 2026 18:00:00 +0000 The safety of large language model (LLM) agents depends not only on model weights but also on the agent harness that manages context, memory, tools, permissions, and runtime control. Existing safety m... http://arxiv.org/abs/2608.09885v1 http://arxiv.org/abs/2608.09885v1#2026-08-11 Tue, 11 Aug 2026 18:00:00 +0000 AI agents increasingly work inside systems that govern how they delegate tasks, move information, execute actions, and use shared resources. Recent work already shows that deployment rules can change ... http://arxiv.org/abs/2608.09828v1 http://arxiv.org/abs/2608.09828v1#2026-08-11 Tue, 11 Aug 2026 18:00:00 +0000