Recent guidance methods in diffusion models steer reverse sampling by perturbing the model to construct an implicit weak model and guide generation away from it. Among these approaches, attention perturbation has demonstrated strong empirical performance in unconditional scenarios where classifier-free guidance is not applicable. However, existing attention perturbation methods lack principled approaches for determining where perturbations should be applied, particularly in Diffusion Transformer (DiT) architectures where quality-relevant computations are distributed across layers. In this paper, we investigate the granularity of attention perturbations, ranging from the layer level down to individual attention heads, and discover that specific heads govern distinct visual concepts such as structure, style, and texture quality. Building on this insight, we propose "HeadHunter", a systematic framework for iteratively selecting attention heads that align with user-centric objectives, enabling fine-grained control over generation quality and visual attributes. In addition, we introduce SoftPAG, which linearly interpolates each selected head's attention map toward an identity matrix, providing a continuous knob to tune perturbation strength and suppress artifacts. Our approach not only mitigates the oversmoothing issues of existing layer-level perturbation but also enables targeted manipulation of specific visual styles through compositional head selection. We validate our method on modern large-scale DiT-based text-to-image models including Stable Diffusion 3 and FLUX.1, demonstrating superior performance in both general quality enhancement and style-specific guidance. Our work provides the first head-level analysis of attention perturbation in diffusion models, uncovering interpretable specialization within attention layers and enabling practical design of effective perturbation strategies.

Along with that new round, it has also launched a new AI-powered bot, Helena, that has a few impressive tricks up its digital sleeve. But the bot goes further than that. It approaches those candidates once they have been discovered, acting as a corporate headhunter. But Helena also completes the recruitment circle by acting as the job seeker's agent. Effectively, it works on behalf of the company and the passive job seeker at the same time, sparing both parties the need to search for each other actively. Powered by machine learning and artificial intelligence, Helena is more than just a bot.

New research and analysis from Federal Reserve economists reveal a problem of mismatches between workers, salaries and productivity, but doesn't identify and discuss the structural cause of the problem: counterproductive recruiting, writes Ask the Headhunter columnist Nick Corcodilos. Nick Corcodilos started headhunting in Silicon Valley in 1979 and has answered over 30,000 questions from the Ask The Headhunter community. In this special Making Sen$e edition of Ask The Headhunter, Nick shares insider advice and contrarian methods about winning and keeping the right job, on one condition: that you, dear Making Sense reader, send Nick your questions about your personal challenges with job hunting, interviewing, networking, resumes, job boards or salary negotiations. No guarantees -- just a promise to do his best to offer useful advice. Recent research by the Federal Reserve suggest that switching jobs -- and probably employers -- is the best way to boost your salary and your career.