AI & ML Papers

🔥 Can LLM Agents Stick to the Script? A Benchmark for Long-Horizon Consistency in Interactive Narratives 💡 The study focuses on the problem of maintaining long-horizon logical consistency in interactive storytelling with large language models. Despite the rapid advancement of large language models, existing research has largely overlooked the challenge of preserving narrative integrity against unconstrained user interventions. To address this, the authors formulate the challenge as Narrative Commitment Preservation and introduce a benchmark called NCP-Bench, which consists of 100 narrative environments derived from movie synopses. Each environment includes a structured narrative specification that can be automatically checked throughout the interaction between the player agent and the narrator agent. The authors use NCP-Bench to evaluate the performance of state-of-the-art large language models in maintaining narrative commitment preservation. The results reveal a substantial long-horizon consistency gap, showing that high linguistic quality does not guarantee commitment preservation. Even strong models frequently generate logically conflicting content under adversarial interventions. The best-performing model achieved only 42 percent survival rate after 20 turns, and fact conflict rates ranged from 40 to 68 percent across models. Furthermore, only isolated runs satisfied all achievement commitments within the 100-turn limit. The study contributes to the field by introducing a new benchmark for evaluating long-horizon consistency in interactive narratives and by highlighting the importance of narrative commitment preservation in interactive storytelling with large language models. The results demonstrate the need for further research to improve the ability of large language models to maintain narrative integrity and consistency in interactive storytelling. Overall, the study provides a comprehensive evaluation of the current state of large language models in interactive storytelling and identifies areas for future improvement. 📅 Published on Aug 8 🔗 Links: • GitHub: https://github.com/huggingface • arXiv: https://arxiv.org/abs/2608.08160 • PDF: https://arxiv.org/pdf/2608.08160 ━━━━━━━━━━━━━━━━━━━━━━━━ 📢 By: https://t.me/PaperNexus #LargeLanguageModels #InteractiveStorytelling #NarrativeConsistency #LongHorizonPlanning #NaturalLanguageProcessing

🔥 M^{2}SNet: Multi-scale in Multi-scale Subtraction Network for Medical Image Segmentation 💡 The paper proposes a new approach to medical image segmentation called M2SNet, which stands for Multi-scale in Multi-scale Subtraction Network. The problem addressed in this paper is that traditional methods for medical image segmentation often result in inaccurate localization and blurred edges of lesions due to the generation of redundant information. This is because most existing methods use element-wise addition or concatenation to fuse different level features, which can weaken the complementarity between these features. To address this challenge, the authors propose a general multi-scale subtraction network that captures detailed and structural cues to improve localization and edge sharpness. The method consists of a basic subtraction unit that produces the difference features between adjacent levels in the encoder. This unit is then expanded to an intra-layer multi-scale subtraction unit that provides both pixel-level and structure-level difference information to the decoder. The authors also pyramidally equip the multi-scale subtraction units at different levels with varying receptive fields to achieve inter-layer multi-scale feature aggregation. Additionally, the authors build a training-free network called LossNet to comprehensively supervise the task-aware features from the bottom layer to the top layer. This drives the multi-scale subtraction network to capture both detailed and structural cues simultaneously. The results show that the proposed method performs favorably against most state-of-the-art methods under different evaluation metrics on eleven datasets of four different medical image segmentation tasks, including color colonoscopy imaging, ultrasound imaging, computed tomography, and optical coherence tomography. The source code is available for further implementation and testing. Overall, the paper contributes a new approach to medical image segmentation that improves localization and edge sharpness by capturing multi-scale difference information. 📅 Published on Mar 20, 2023 🔗 Links: • GitHub: https://github.com/huggingface • arXiv: https://arxiv.org/abs/2303.10894 • PDF: https://arxiv.org/pdf/2303.10894 ━━━━━━━━━━━━━━━━━━━━━━━━ 📢 By: https://t.me/PaperNexus #MedicalImageSegmentation #MultiScaleSubtractionNetwork #M2SNet #DeepLearningForMedicalImaging #ImageSegmentationTechniques

🔥 Spark-to-Paper: End-to-End Research Paper Generation as a Composable Skill 💡 The paper presents Spark-to-Paper, a system that generates research papers from start to finish, using a composable workflow inside coding assistants. The problem addressed is that generating a research paper requires more than just text generation, it involves retrieving literature, designing and executing experiments, revising claims based on evidence, producing publication-ready figures, and maintaining consistency throughout the process. The method used by Spark-to-Paper separates planning from reporting, and enforces evidence-based claim revision. It combines deterministic integrity checks with self-critique to improve reliability and prevent a failure mode called the Self-Refutation Loop, where repeated experiments reject the original research objective. The system also produces editable vector figures through programmatic plotting for experimental results and code-based reconstruction for generated method diagrams. The results show that Spark-to-Paper achieves high citation validity and figure editability, with 99.5 percent citation validity and 96.4 percent figure editability across eight controlled research topics. The system also detects fabrication effectively, increasing detection from 14 percent for a single-pass draft to 92 percent with the full integrity and review stack. The full system is efficient, using 11.9 million tokens, costing 8.1 dollars per manuscript, and requiring 3.2 hours on average. Overall, the paper demonstrates that end-to-end research paper generation can be implemented as a lightweight and composable workflow inside existing coding assistants, keeping experimental evidence central to the research process. 📅 Published on Aug 12 🔗 Links: • GitHub: https://github.com/huggingface • arXiv: https://arxiv.org/abs/2608.11924 • PDF: https://arxiv.org/pdf/2608.11924 ━━━━━━━━━━━━━━━━━━━━━━━━ 📢 By: https://t.me/PaperNexus #ResearchPaperGeneration #ComposableWorkflow #AutomatedScientificWriting #EndToEndResearch #AIassistedAcademicWriting

🚀 21-Day CCNA & CCNP Sprint – Aug 17 to Sep 6 🤝 Peer Group – Share Insights | Exchange Knowledge | Support Each Other No more studying alone. Join a community of CCNA/CCNP candidates, learn together, and win prizes. How it works: ① DM admin: "I'M IN + cert name" ② Join the group ③ Check in 18/21 days → win 🎁 Prizes (first come, first served): $50 Amazon card ×1 | SD-Access Training ×1 | SD-WAN Training ×1 | CCNA Pro Package ×10 | Free Learning Pack (all finishers) Daily check-in: 1️⃣ What you learned 2️⃣ Explain it in your own words 3️⃣ (Optional) Ask the group Join now: https://chat.whatsapp.com/KZrAj2HZ3Y5K9UhhNhrApf DM to register: https://wa.me/8619559123054

🔥 Co-Evolution in Agentic Systems: Toward Self-Directed Evolution Beyond Human Design 💡 The paper discusses the concept of co-evolution in agentic systems, which allows for open-ended improvement beyond human design. The problem addressed is that single-entity self-evolution in agentic systems is often limited by a static learning context, such as fixed tasks and feedback. To overcome this, the paper proposes a multi-component co-evolution approach, where multiple agents and their environment adapt and impose pressure on each other. The method involves a three-stage taxonomy to organize existing research on co-evolution in agentic systems. The first stage, Agent-Agent Co-Evolution, focuses on how agents adapt through dynamic interactions with peers, including adversarial, collaborative, and organizational adaptation. The second stage, Agent-Environment Co-Evolution, extends this concept to adaptive tasks, feedback, and interaction spaces that change with the agents. The third stage, Meta Co-Evolution, explores the possibility of making the evolution mechanism itself evolvable. The results of this approach are agentic systems that can improve beyond fixed human-designed paths. The paper provides a unified foundation for building robust and open-ended agentic systems, and discusses open challenges in evaluating, scaling, and controlling such systems. The contributions of the paper include a comprehensive survey of co-evolution in agentic systems, a progressive taxonomy to organize existing research, and a framework for building self-directed evolutionary systems that can adapt and improve over time. Overall, the paper provides a new perspective on achieving open-ended improvement in agentic systems, and highlights the potential of co-evolution as a key mechanism for creating more autonomous and adaptive systems. 📅 Published on Aug 10 🔗 Links: • GitHub: https://github.com/huggingface • arXiv: https://arxiv.org/abs/2608.10299 • PDF: https://arxiv.org/pdf/2608.10299 ━━━━━━━━━━━━━━━━━━━━━━━━ 📢 By: https://t.me/PaperNexus #CoEvolutionInAgenticSystems #SelfDirectedEvolution #AgenticSystemsDesign #MultiComponentCoEvolution #OpenEndedIntelligence

🔥 Business Arena: Benchmarking LLM Agents in a Realistic Marketplace 💡 The paper introduces Business Arena, a benchmarking environment that evaluates the performance of large language model agents in running a realistic cross border shop. The goal is to assess the ability of these agents to make business decisions and operate a business in a challenging and dynamic market. The environment is grounded in real data from Alibaba.com and market conditions, and agents are tasked with buying from suppliers and selling to buyers over a long period of time. The performance of the agents is measured by their final net worth, and the results show a significant gap between the best performing agent and human designed strategies. The paper also uses skill level metrics and action level attribution to analyze the strengths and weaknesses of the agents and identify the decisions that lead to success or failure. The results show that even the best performing agent falls behind human designed strategies, indicating that operating a business remains a challenging task for large language model agents. The paper evaluates 15 frontier models and finds a ninefold difference in mean final net worth, with the best model still falling short of human performance. The analysis also reveals different operating styles among the agents, including margin focused premium sellers, high turnover wholesalers, and customer service specialists. Overall, the paper contributes to the development of a realistic and trustworthy testbed for evaluating end to end business agents, and highlights the need for further research to improve the performance of large language model agents in business operations. 📅 Published on Aug 9 🔗 Links: • GitHub: https://github.com/huggingface • arXiv: https://arxiv.org/abs/2608.08621 • PDF: https://arxiv.org/pdf/2608.08621 • Project Page: https://business-arena.site.accio.ai/ ━━━━━━━━━━━━━━━━━━━━━━━━ 📢 By: https://t.me/PaperNexus #LLMAgents #BusinessDecisionMaking #CrossBorderTrade #MarketplaceSimulation #LargeLanguageModels