@anik

Multi-agent systems achieve state-of-the-art outcomes through peer collaboration. However, when an agent in the pipeline silently drops a constraint, the system's final output may look correct even though the reasoning chain was quietly corrupted, and existing outcome-based evaluations are blind to such multi-hop process failures. To make these vulnerabilities measurable before deployment, we introduce AgentCollabBench, a diagnostic benchmark of 900 human-validated tasks spanning software engineering, DevOps, and data engineering. Each task isolates one of four behavioral risks: instruction decay (does a constraint survive peer pressure?), false-belief contagion (does a falsehood spread through consensus?), context leakage (does information bleed between tasks?), and tracer durability (does marked data reach the final agent?). Evaluating four modern LLMs (GPT 4.1 mini, Gemini 2.5 Flash Lite, Qwen-3.5-35B-A3B, and Llama 3.1 8B Instruct), we expose model-specific vulnerability profiles invisible to outcome-only evaluation; Qwen-3.5-35B-A3B, for example, leads on tracer durability and instruction stability, while GPT 4.1 mini leads on leakage containment and false-belief resistance. Beyond per-model differences, communication topology emerges as a primary risk factor that explains 7-40% of the variance in multi-hop information survival. The effect traces to a synthesis bottleneck specific to converging-DAG nodes: an agent weighing competing parent inputs discards constraints carried by a minority branch, a bottleneck structurally absent from linear chains. AgentCollabBench demonstrates that suboptimal topology can silently erase the safeguards of highly capable models, arguing that multi-agent reliability is fundamentally a structural problem and that scaling model intelligence alone is no substitute for architecture.

Understanding who plays which narrative role in a meme requires jointly interpreting the image, embedded text, and culturally specific context. This is particularly difficult for Bengali memes, where resources and annotated benchmarks are limited and most prior work targets coarse tasks such as hate speech or sentiment detection rather than entity-level narrative roles. Bengali meme understanding is formulated as an entity-centric role labeling and explanation task. To support this setting, a dataset of 2,658 Bengali memes is curated and annotated with role labels (hero, villain, victim, and other), along with context notes used to resolve culture-dependent references. Two complementary multimodal solutions are evaluated: (i) a triple-stream fusion classifier that integrates visual, textual, and context signals via attention-based fusion, and (ii) a vision–language reasoning pipeline that performs holistic inference over full meme content and produces explanations. Methodologically, the work contributes an entity-centric benchmark with context annotations and evaluates attention-based fusion and vision–language reasoning for role labeling and grounded explanation generation. Practically, it enables culturally grounded analysis of Bengali memes for downstream understanding of satire and social commentary in real-world social-media content. Experimental results show that the vision–language pipeline improves over the fusion baseline, achieving 0.8068 accuracy and 0.7953 macro-averaged F1. In addition, the explanation generator improves rubric-based quality from 6.03/10 to 7.33/10 under a multi-model judging setup.