AI coding assistants powered by large language models (LLMs) have transformed software development, significantly boosting productivity. While existing benchmarks evaluate the correctness and security of LLM-generated code, they are limited to single-turn tasks that cannot capture the conversational nature of real-world development, where requirements evolve across turns and tools increasingly generate incremental diffs rather than complete programs. We introduce MT-Sec, the first benchmark to jointly evaluate correctness and security in multi-turn coding scenarios. MT-Sec covers four interaction types (expansion, editing, refactoring, and debugging) reflecting common development workflows. We construct multi-turn tasks using human-curated instructions and feedback-driven interactions grounded in dynamic test suites. Evaluating 31 LLMs and 3 agent scaffolds, we find that models struggle most with expansion and editing, but improve significantly in debugging when provided with feedback. Agent-based systems often underperform their base models in automated multi-turn settings due to brittle tool use. We also evaluate multi-turn code-diff generation, an unexplored but practical setting, and observe higher failure rates compared to full-program generation. Our findings underscore the need for multi-turn benchmarks to drive safer AI-assisted development.

Automatically generated passwords and passphrases are a cornerstone of IT security. Yet, these passphrases are often hard to remember and see only limited adoption. In this work, we use large language models to generate passphrases with rigorous security guarantees via the computation of the entropy of the output as a metric of the security of the passphrase. We then present a range of practical methods to generate language model outputs with sufficient entropy: raising entropy through in-context examples and generation through a new top-q truncation method. We further verify the influence of prompt construction in steering the output topic and grammatical structure. Finally, we conduct user studies to determine the adoption rates for these LLM-generated passphrases in practice.

Large vision-language models (LVLMs) have achieved impressive results in visual question-answering and reasoning tasks through vision instruction tuning on specific datasets. However, there remains significant room for improvement in aligning visual and language modalities. Existing methods often depend on external models or data, leading to uncontrollable and unstable alignment results. In this paper, we propose SIMA, a self-improvement framework that enhances visual and language modality alignment without external dependencies. SIMA leverages existing vision instruction tuning datasets to self-generate responses, incorporating an in-context self-critic mechanism that constructs preference pairs for tuning. Crucially, our approach allows LVLMs to act as critics by designing effective critic prompts, eliminating the need for additional fine-tuning with external instruction data. We introduce three novel visual metrics within the self-critic process to guide judgement, significantly improving the accuracy of self-critic. Through extensive experiments across 14 hallucination and comprehensive benchmarks, we demonstrate that SIMA significantly improves LVLM’s performance and outperforms previous approaches, achieving superior modality alignment.