Every paper checked — nothing hidden
All papers
Every paper checked for brief #4 of 12–19 Oct 2026 — nothing is hidden. Brief papers first.
All17,312In brief24More (on-topic)3,333Not relevant13,753Not assessed87Older351Could not confirm93Blogs with no paper266All blogs284
| Paper | Topic | Relevance | Feedback |
|---|---|---|---|
| Demystifying On-Policy Distillation: Where Does the Teacher's Advantage Live? The paper examines teacher interventions along student generation paths to identify how teacher guidance drives performance gains in on-policy distillation. Ranked #158Surprising resultPreprint6/10 suggested | Small and efficient models via arxiv · 15 Oct, 15:06 | On-topic Analyzes on-policy distillation mechanisms to understand how student models effectively absorb teacher knowledge. | |
| EvoSpatial: Experience-Driven Skill and Tool Evolution for Spatial Reasoning Agents The paper proposes a framework that evolves a reusable library of skills and tools based on experience for spatial reasoning agents. Ranked #158New way to see a problemPreprint6/10 suggested | Multimodal models via arxiv · 15 Oct, 06:18 | On-topic Enhances tool-augmented vision-language agents performing multimodal spatial reasoning. | |
| Instructions Are Evidence, Not Objectives: Measuring Outcome Alignment of Agents This paper introduces an approach and benchmark for evaluating whether language model agents can infer unstated true requirements instead of following incomplete instructions. Ranked #158New way to see a problemPreprint7/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | On-topic Introduces TiGanEval, a benchmark measuring whether agents discover true requirements beyond stated instructions. | |
| ProtoArena: Can AI Agents Discover and Exploit Real-World Cross-Component Protocol Vulnerabilities? Introduces ProtoArena, a black-box dynamic benchmark evaluating the ability of AI agents to discover and exploit cross-component network protocol vulnerabilities. Ranked #158New testPreprint7/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Evaluates autonomous LLM agents acting across environments to find and exploit protocol vulnerabilities. | |
| Same Outcome, Different Evidence: Intent Recovery in LLM Safety Evaluation Pairs attack success rate with operative understanding rate to assess whether models recognize harmful intent during safety evaluations. Ranked #158Surprising resultPreprint7/10 suggested | AI safety and reliability in deployment via arxiv · 12 Oct, 16:28 | On-topic Investigates jailbreak evaluation metrics, harmful output detection, and safety assessment validity. | |
| Task-Phase-Conditioned Expert Routing Is Not a Reliable Proxy for Quantization Sensitivity in MoE FFN Tensors This paper shows that routing statistics across task phases fail to provide an effective guide for mixed-precision quantization in mixture-of-experts models. Ranked #159Surprising resultPreprint6/10 suggested | Small and efficient models via arxiv · 17 Oct, 06:18 | On-topic Studies mixed-precision quantization strategies for Mixture-of-Experts language models. | |
| Detecting Chain-of-Thought Hijacking in Backdoored Multimodal Large Reasoning Models The paper introduces a method to detect adversarial chain-of-thought hijacking in backdoored multimodal reasoning models by monitoring reasoning dynamics. Ranked #159New way to see a problemPreprint6/10 suggested | Multimodal models via arxiv · 15 Oct, 06:18 | On-topic Investigates chain-of-thought reasoning security in multimodal large reasoning models. | |
| When Should Enterprise Policy Decisions Be Learned, and When Should They Be Reasoned? Evaluates enterprise policy enforcement across structured JSON versus single-token outputs, measuring the impact of explicit reasoning on decision stability. Ranked #159New way to see a problemPreprint7/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Studies action selection mechanisms and tool calling in enterprise policy agents. | |
| Auditing Failure Attribution: Label Predictability and Judge Behavior This paper provides four auditing tests to determine whether failure-attribution judges actually perform multi-step checks or merely rely on spurious surface cues. Ranked #159Surprising resultPreprint7/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | On-topic Exposes how surface cues distort LLM-as-a-judge failure attribution benchmarks and proposes audit tests. | |
| Do Prompt Optimizers Break Requirements They Are Not Scored On? A Preregistered Study Evaluates whether automated prompt optimizers inadvertently cause regressions in unscored prompt constraints and requirements. Ranked #159Surprising resultPreprint7/10 suggested | AI safety and reliability in deployment via arxiv · 12 Oct, 16:28 | On-topic Evaluates deployment reliability and unintended side effects or failures under prompt optimization. | |
| Teach Me After You Learn It: Lookahead Self-Distillation for RLVR This paper proposes Lookahead Self-Distillation, providing cross-instance token-level supervision to improve reinforcement learning with verifiable rewards for language models. Ranked #160New way to see a problemPreprint6/10 suggested | Small and efficient models via arxiv · 15 Oct, 06:18 | On-topic Investigates on-policy self-distillation techniques to train language models more effectively. | |
| Self-Verifier Capture in Unified Visual Generators: Diagnosing and Repairing Closed-Loop Self-Reward The paper analyzes failure modes in unified visual generators using self-reward loops and proposes stabilization methods to prevent quality degradation. Ranked #160New way to see a problemPreprint6/10 suggested | Multimodal models via arxiv · 15 Oct, 06:18 | On-topic Studies self-reward alignment in unified visual generators with joint understanding pathways. | |
| How Safe Is ChatGPT Health Advice? Evidence from Real User Conversations in Four Global South Countries Evaluates the safety of ChatGPT health advice using real user conversation logs across four Global South countries. Ranked #160Surprising resultPreprint7/10 suggested | AI safety and reliability in deployment via arxiv · 15 Oct, 06:17 | On-topic Evaluates real-world deployment risks, unsafe health guidance, and harmful recommendations in ChatGPT. | |
| Agency as a New Organizing Abstraction for Computer Science The paper argues that software agency should be treated as a core organizing abstraction in computer science rather than merely an AI application. Ranked #160New way to see a problemPreprint7/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Explores agentic AI systems that synthesize actions across tools and open environments. | |
| Visual Dependence Does Not Imply Answerability: A Systematic Analysis of Answerability Signals in Frozen InstructBLIP Analyzes answerability signals in frozen visual question answering models, finding standard confidence and alignment metrics fail to reliably detect unanswerable inputs. Ranked #160Surprising resultPreprint7/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | On-topic Provides a systematic evaluation of answerability signals and unanswerable question discrimination in VLMs. | |
| Genesis: Visual Augmentation Primitives for Multimodal Chain-of-Thought Reasoning The paper introduces an image augmentation pipeline that generates diverse visual variants to anchor multimodal chain-of-thought reasoning in visual evidence. Ranked #161New way to see a problemPreprint6/10 suggested | Multimodal models via arxiv · 15 Oct, 06:18 | On-topic Introduces visual primitives to anchor multimodal chain-of-thought reasoning in vision-language models. | |
| Get In, Decide, Prove: A Merchant-Side Readiness Benchmark for Agentic Commerce The paper introduces a benchmark assessing whether merchant websites provide adequate access, terms, and evidence for autonomous shopping agents. Ranked #161New testPreprint7/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Examines merchant readiness and interaction bottlenecks for autonomous buying AI agents. | |
| Efficiently Learning To Reason or Not to Reason: Root-token Policy Optimization for Adaptive Thinking This paper presents an optimization method that trains models to decide when to invoke reasoning tokens, reducing unnecessary thinking overhead. Ranked #161Surprising resultPeer-reviewed6/10 suggested | Small and efficient models via arxiv · 15 Oct, 06:17 | On-topic Reduces training compute and VRAM overhead by adaptively skipping unnecessary reasoning traces. | |
| Interaction-as-a-Judge Presents Interaction-as-a-Judge, using computer-use agents to evaluate interactive web and desktop software applications through direct dynamic interaction. Ranked #161New testPreprint7/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | On-topic Proposes an interactive evaluation paradigm and releases benchmarks for judging interactive applications. | |
| SOAR: Source-Aware Reinforcement Learning with Precise Credit Assignment for Search Agents Proposes a reinforcement learning framework that tracks external versus internal knowledge sources to reduce search agent fabrication. Ranked #161New way to see a problemPreprint7/10 suggested | AI safety and reliability in deployment via arxiv · 12 Oct, 16:28 | On-topic Directly tackles evidence fabrication and hallucination during multi-step reasoning. | |
| DRanker: A Transformer-Based Multi-Task Ranking Model for the Disney+ Homepage The paper presents an encoder-only transformer ranking model deployed at Disney+ that uses lightweight debiasing to improve recommendations. Ranked #162Surprising resultPeer-reviewed6/10 suggested | Small and efficient models via arxiv · 15 Oct, 06:18 | On-topic Details an efficient transformer architecture and lightweight inference stack for production ranking. | |
| VideoEvoHarness: Self Evolving Agent Harness from Unlabeled Videos for Video Understanding The paper proposes an agent harness that autonomously evolves questioner and solver workflows from unlabeled video data without updating base model parameters. Ranked #162New way to see a problemPreprint6/10 suggested | Multimodal models via arxiv · 15 Oct, 06:18 | On-topic Develops an evolving agent harness specifically for multimodal video understanding. | |
| Upstream Influence, Downstream Re-Anchoring in LLM Agent Chains Analyzes how multi-agent workflow architecture affects the propagation and attenuation of coordinated stance manipulation across sequential agent chains. Ranked #162Surprising resultPreprint7/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Analyzes how manipulation propagates through multi-agent LLM chains. | |
| ProtoArena: Can AI Agents Discover and Exploit Real-World Cross-Component Protocol Vulnerabilities? Introduces ProtoArena, a black-box dynamic benchmark evaluating the ability of AI agents to discover and exploit cross-component network protocol vulnerabilities. Ranked #162New testPreprint7/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | On-topic Introduces a dynamic benchmark evaluating agent capabilities in exploiting network protocol vulnerabilities. | |
| SteerablePlex: Can We Steer Full-Duplex Models? Introduces an instruction-following benchmark and training recipe to improve goal adherence in full-duplex conversational models. Ranked #162New testPreprint7/10 suggested | AI safety and reliability in deployment via arxiv · 12 Oct, 16:28 | On-topic Addresses steerability, constraint compliance, and reliability of conversational model behavior. | |
| Dissecting the Carbon-SLO Tension in Geo-Distributed LLM Serving The paper evaluates carbon-aware scheduling for geographically distributed language model serving under strict latency service-level objectives using an empirical simulator. Ranked #163New way to see a problemPreprint6/10 suggested | Small and efficient models via arxiv · 15 Oct, 06:18 | On-topic Investigates latency and scheduling optimizations for efficient geo-distributed LLM serving. | |
| TempoKV: Predictive KV Cache Quantization for Autoregressive Video Generation The paper presents a predictive KV cache quantization method exploiting temporal predictability to reduce memory overhead in autoregressive video generation models. Ranked #163New way to see a problemPreprint6/10 suggested | Multimodal models via arxiv · 15 Oct, 06:18 | On-topic Provides predictive KV cache quantization for autoregressive multimodal video generation models. | |
| LatentRiskBench: Measuring State-Dependent Risks in Enterprise Agents Introduces LatentRiskBench to evaluate safety risks in enterprise agents where action safety depends on latent workflow context. Ranked #163New testPreprint7/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Evaluates multi-step enterprise agent workflows and state-dependent policy violations. | |
| SWEPatchTest: Evaluating Software Tests Beyond Reproduction SWEPatchTest provides a benchmark to evaluate whether agent-generated test suites correctly accept valid patches and reject invalid ones. Ranked #163New testPreprint7/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | On-topic Introduces a new evaluation benchmark for software tests generated for code patches. | |
| Could LLM Watermark Detection be Public? Evaluates public LLM watermark detectors and introduces a split-key mechanism combining public detection with private verification. Ranked #163New way to see a problemPreprint7/10 suggested | AI safety and reliability in deployment via arxiv · 12 Oct, 16:28 | On-topic Studies LLM output watermarking, detector exposure liabilities, and tamper-resistance for safe deployment. | |
| VisualCoding: Learning Executable Skills for Reliable Visual Reasoning This paper introduces a framework that acquires and directly executes reusable visual programs to improve multimodal reasoning agent reliability. Ranked #164New way to see a problemPreprint6/10 suggested | Multimodal models via arxiv · 15 Oct, 06:18 | On-topic Combines executable code acquisition with visual reasoning to enhance multimodal question answering. | |
| Glasseek: A Transparent Recipe for Training and Trustworthy Evaluation of DeepSearch Agents The paper introduces an open-source framework and recipe for training and evaluating deep-search language agents on complex web exploration tasks. Ranked #164New testPreprint7/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Presents Glasseek, an open framework for training and executing deep web-search agents. | |
| Large Language Models Extend Known Research But Do Not Create New Research Directions The study evaluates scientific hypotheses generated by language models, concluding that they extend existing knowledge rather than starting new research directions. Ranked #164Surprising resultPreprint7/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | On-topic Proposes a framework to evaluate whether LLMs generate genuinely novel scientific hypotheses. | |
| ChunkAct: Action Chunking for Agentic Robotics The paper introduces action chunking for agentic robotics to execute pre-planned primitive routines without redundant model reasoning at each step. Ranked #164New way to see a problemPreprint6/10 suggested | Small and efficient models via arxiv · 15 Oct, 06:17 | On-topic Reduces LLM calls and reasoning costs during multi-step robotic agent execution. | |
| A Security Meta-Model for Retrieval-Augmented Generation Systems Proposes a security meta-model mapping attack surfaces, vulnerabilities, and risks across retrieval-augmented generation pipelines. Ranked #164New way to see a problemPreprint7/10 suggested | AI safety and reliability in deployment via arxiv · 12 Oct, 16:28 | On-topic Proposes a security meta-model mapping vulnerabilities, attack surfaces, and risks across RAG deployments. | |
| Decoupling Knowledge Forgetting and Refusal Learning in MLLM Unlearning The paper proposes a distillation framework to remove specific multimodal knowledge without causing hallucinations or leaving residual queryable information. Ranked #165New way to see a problemPreprint6/10 suggested | Multimodal models via arxiv · 15 Oct, 06:17 | On-topic Addresses multimodal knowledge forgetting and refusal learning specifically in multimodal LLMs. | |
| Audit tools of the trade: Automated Auditor Elicitation Can Be Sensitive to the Context the Auditor Is Given The paper examines how providing additional context, such as codebases and benchmark transcripts, affects automated LLM safety auditors. Ranked #165Surprising resultPreprint7/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Studies automated LLM auditors evaluating coding agents across diverse contexts. | |
| Get In, Decide, Prove: A Merchant-Side Readiness Benchmark for Agentic Commerce The paper introduces a benchmark assessing whether merchant websites provide adequate access, terms, and evidence for autonomous shopping agents. Ranked #165New testPreprint7/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | On-topic Introduces a benchmark evaluating merchant sites for agentic decision sufficiency. | |
| Lightweight but Linguistically Fair? Quantisation Robustness across Bavarian, German, and English The authors evaluate how 8-bit and 4-bit quantization methods affect model accuracy across Bavarian, Standard German, and English benchmarks. Ranked #165Surprising resultPreprint6/10 suggested | Small and efficient models via arxiv · 15 Oct, 06:17 | On-topic Evaluates the impact of 8-bit and 4-bit quantization methods across dialects. | |
| Understanding and Mitigating Cross-Session Attacks via Persistent Artifacts in Agent Skill Composition The paper analyzes vulnerabilities from persistent artifacts shared across AI agent skill sessions and develops a runtime defense against resulting cross-session attacks. Ranked #165New testPreprint7/10 suggested | AI safety and reliability in deployment via arxiv · 12 Oct, 16:28 | On-topic Studies cross-session injection and poisoning attacks in deployed agent architectures. | |
| HumanPhys: A Human-Centric Physical-World Perception Benchmark for Real-World Visual Reasoning Introduces a benchmark probing physical perception tasks related to humans, including depth, trajectory, illumination, and kinematics. Ranked #166New testPreprint6/10 suggested | Multimodal models via arxiv · 15 Oct, 06:18 | On-topic Introduces a benchmark evaluating multimodal LLMs on physical-world visual reasoning. | |
| When Goals Leak Across Relationships: How Cached Goal Representations Shape Cross-Client Decisions The paper examines how shared KV caches cause language model agents to leak goals across competing clients, offering internal intervention techniques. Ranked #166Surprising resultPreprint7/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Analyzes autonomous multi-client agent decision-making and cross-context goal interference. | |
| LatentRiskBench: Measuring State-Dependent Risks in Enterprise Agents Introduces LatentRiskBench to evaluate safety risks in enterprise agents where action safety depends on latent workflow context. Ranked #166New testPreprint7/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | On-topic Introduces LatentRiskBench to evaluate state-dependent enterprise agent safety. | |
| LoOPSD: On-Policy Self-Distillation with Looped Teachers This paper introduces a self-distillation method that searches over repeated layer execution orders to create stronger teacher guidance. Ranked #166New way to see a problemPreprint6/10 suggested | Small and efficient models via arxiv · 15 Oct, 06:17 | On-topic Uses self-distillation across looped layers to improve small model reasoning efficiency. | |
| Vulnerable Models in the AI Supply Chain: An Empirical Study of Adoption, Propagation, and Usage Patterns The paper conducts an empirical study tracking how publicly flagged vulnerable AI models propagate across code repositories and downstream applications. Ranked #166New testPreprint7/10 suggested | AI safety and reliability in deployment via arxiv · 12 Oct, 16:28 | On-topic Investigates the propagation and deployment of vulnerable, backdoored AI models across the supply chain. | |
| VARI: Robust Text-to-Video Generation via Group-Structured Intent Alignment Introduces VARI, a group-structured intent alignment framework to improve text-to-video generation robustness across equivalent prompts. Ranked #167New way to see a problemPreprint6/10 suggested | Multimodal models via arxiv · 15 Oct, 06:18 | On-topic Presents an alignment method for text-to-video multimodal generative models. | |
| Structural Baselines Matter for Agent Failure Detection and Attribution Shows agent failure detection models often rely on trivial cues like run length, exposing benchmark flaws when attributing critical failure steps. Ranked #167Surprising resultPreprint7/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Analyzes failure detection and attribution methods in tool-using LLM agent execution runs. | |
| Glasseek: A Transparent Recipe for Training and Trustworthy Evaluation of DeepSearch Agents The paper introduces an open-source framework and recipe for training and evaluating deep-search language agents on complex web exploration tasks. Ranked #167New testPreprint7/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | On-topic Audits evaluation contamination and establishes trustworthy evaluation practices for search agents. | |
| PaceReasoner: Label-Free Learning to Answer Briefly from Long Reasoning Introduces PaceReasoner, using long reasoning outputs during test-time reinforcement learning to boost the accuracy of brief, low-latency model answers. Ranked #167New way to see a problemPreprint6/10 suggested | Small and efficient models via arxiv · 15 Oct, 06:17 | On-topic Improves short-form reasoning accuracy while drastically reducing token overhead and inference latency. | |
| A persistent accuracy ceiling in automated verbal deception detection The paper meta-analyzes 25 years of automated verbal deception detection models and identifies a persistent accuracy ceiling around 70 to 75 percent. Ranked #167Surprising resultPreprint7/10 suggested | AI safety and reliability in deployment via arxiv · 12 Oct, 16:28 | On-topic Exposes reliability and evaluation validity limitations in deployed verbal deception detection systems. | |
| Diagnose-to-Generate: Eliminating Structural Distortions in Flow Models via Patch-Level Grounded Alignment This paper introduces a grounded alignment framework using anomaly detection datasets and patch-level policy optimization to fix distortions in flow generators. Ranked #168New way to see a problemPreprint6/10 suggested | Multimodal models via arxiv · 15 Oct, 06:18 | On-topic Aligns visual anomaly diagnosis with generative flow-based vision models. | |
| Fine-Tuning Strategy, Not Just Data: A Statistically Rigorous Audit of Gender and Stereotypical Bias in Embedding Models for HR Recommendation Systems This paper audits gender and stereotypical bias in multilingual job-matching embedding models using robust statistical tests. Ranked #168Surprising resultPreprint7/10 suggested | AI safety and reliability in deployment via arxiv · 15 Oct, 06:18 | On-topic Audits social bias and fairness risks in production recommendation deployments. | |
| LATCH: LLM-Assisted Timing Closure Harness for Post-Route FPGA Optimisation Evaluates different agent control granularities for post-route FPGA timing closure to balance execution reliability and frequency gains. Ranked #168Surprising resultPreprint7/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Evaluates LLM agent orchestration, action control granularities, and exploration strategies for FPGA optimization. | |
| Comparing GenAI to Machine Learning and Manual Review for eDiscovery The paper analyzes legal eDiscovery review data to compare generative AI tools against continuous active machine learning and manual human review. Ranked #168Surprising resultPreprint7/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | On-topic Critiques evaluation methodology and compares performance metrics between GenAI and ML. | |
| RT-J: Large-Scale Pretraining of Relational Transformers for Context-Efficient Predictions Presents RT-J, a relational transformer pretrained on diverse databases to achieve strong in-context table predictions with few sample examples. Ranked #168New way to see a problemPreprint6/10 suggested | Small and efficient models via arxiv · 15 Oct, 06:17 | On-topic Develops a small 85M relational model that achieves high accuracy with few in-context examples. | |
| TableVision: Recovering Verified Structure from Historical Tables TableVision recovers verified table structure from historical documents using dynamic programming and column sum checks. Ranked #169New way to see a problemPreprint6/10 suggested | Multimodal models via arxiv · 15 Oct, 06:18 | On-topic Recovers structured data from scanned document images and tables. | |
| From Caricature to Collective Artifact: How Persona Construction Drives Polarization in Generative Social Simulations Analyzes how persona construction in generative social simulations causes network segregation and collective political polarization. Ranked #169Surprising resultPreprint7/10 suggested | AI agents via arxiv · 15 Oct, 06:17 | On-topic Studies collective behavior and emergent polarization across interacting multi-agent LLM personas. | |
| Audit tools of the trade: Automated Auditor Elicitation Can Be Sensitive to the Context the Auditor Is Given The paper examines how providing additional context, such as codebases and benchmark transcripts, affects automated LLM safety auditors. Ranked #169Surprising resultPreprint7/10 suggested | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:17 | On-topic Evaluates how context shapes automated LLM auditor benchmark assessments. | |
| Social Priors: Descriptions of AI Users Shape How Language Models Trust and Cooperate The paper investigates how fine-tuning language models on descriptions of user motivations influences their downstream trust and cooperative behavior. Ranked #169Surprising resultPreprint7/10 suggested | AI safety and reliability in deployment via arxiv · 15 Oct, 06:17 | On-topic Analyzes how training data narratives affect model trust, cooperation, and safety risks. | |
| LatentRemedy: Efficient Remedy For Non-Prefix KV Cache Reuse in Multi-Agent Systems LatentRemedy uses task-conditioned latent tokens to allow safe key-value cache reuse across conflicting tasks in multi-agent systems. Ranked #169New way to see a problemPreprint6/10 suggested | Small and efficient models via arxiv · 15 Oct, 06:17 | On-topic Improves inference efficiency by enabling non-prefix KV cache reuse across agents. |
1,081–1,140 of 3,333← PreviousNext →