Constructing inputs that drive execution to chosen code locations underpins many dynamic-analysis tasks in software engineering and security. It is a generalized constraint-solving challenge that requires agents to reason about the semantic mapping between input and code. We built COIN, the first benchmark to systematically evaluate this ability in modern code agents. By leveraging OSS-Fuzz’s long-running coverage data and maintainer-written harnesses, COIN constructs reachability challenges across over 1000 real-world projects in nine languages, and is contamination-resistant and self-refreshing as upstream code and OSS-Fuzz coverage evolve. To keep challenges non-trivial yet solvable, COIN samples two target families: lines OSS-Fuzz has covered but baselines miss, and lines never be reached but with static evidence of a reachable harness path. Across five independent runs on 137 targets from 15 projects in six languages, the strongest agent (Claude Code with Opus 4.6) reaches on average only 26.6% of the C/C++ targets and 39.7% of the Go/JVM/Python/Rust targets, never half of either set in any run, and the weakest (DeepSeek V3.2) stays below 9%, indicating substantial room for improvement.

Evaluating text generation capabilities of large language models (LLMs) is challenging, particularly for low-resource languages where methods for direct assessment are scarce. We propose MUG-Eval, a novel framework that evaluates LLMs’ multilingual generation capabilities by transforming existing benchmarks into conversational tasks and measuring the LLMs’ accuracies on those tasks. We specifically designed these conversational tasks to require effective communication in the target language. Then, we simply use task success rate as a proxy for successful conversation generation. Our approach offers two key advantages: it is independent of language-specific NLP tools or annotated datasets, which are limited for most languages, and it does not rely on LLMs-as-judges, whose evaluation quality degrades outside a few high-resource languages. We evaluate 8 LLMs across 30 languages spanning high, mid, and low-resource categories, and we find that MUG-Eval correlates strongly with established benchmarks (r > 0.75) while enabling standardized comparisons across languages and models. Our framework provides a robust and resource-efficient solution for evaluating multilingual generation that can be extended to thousands of languages.

We introduce LLM-as-an-Interviewer, a novel paradigm for evaluating large language models (LLMs). This approach leverages multi-turn interactions where the LLM interviewer actively provides feedback on responses and poses follow-up questions to the evaluated LLM. At the start of the interview, the LLM interviewer dynamically modifies datasets to generate initial questions, mitigating data contamination. We apply the LLM-as-an-Interviewer framework to evaluate six models on the reasoning, factuality and instruction-following tasks. Our results show that the framework effectively provides insights into LLM performance, including the quality of initial responses, adaptability to feedback, and ability to address follow-up queries like clarification or additional knowledge requests. The framework also addresses key limitations of conventional methods like LLM-as-a-Judge, including verbosity bias and inconsistency across runs. Finally, we propose the Interview Report, which aggregates insights from the interview process, providing examples and a comprehensive analysis of the LLM’s strengths and weaknesses. This report offers a detailed snapshot of the model’s real-world applicability.

While humans naturally develop theory of mind (ToM), the capability to understand other people’s mental states and beliefs, state-of-the-art large language models (LLMs) underperform on simple ToM benchmarks. We posit that we can extend our understanding of LLMs’ ToM abilities by evaluating key human ToM precursors-perception inference and perception-to-belief inference-in LLMs. We introduce two datasets, Percept-ToMi and Percept-FANToM, to evaluate these precursory inferences for ToM in LLMs by annotating characters’ perceptions on ToMi and FANToM, respectively.Our evaluation of eight state-of-the-art LLMs reveals that the models generally perform well in perception inference while exhibiting limited capability in perception-to-belief inference (e.g., lack of inhibitory control).Based on these results, we present PercepToM, a novel ToM method leveraging LLMs’ strong perception inference capability while supplementing their limited perception-to-belief inference. Experimental results demonstrate that PercepToM significantly enhances LLM’s performance, especially in false belief scenarios.

Additive compositionality of word embedding models has been studied from empirical and theoretical perspectives. Existing research on justifying additive compositionality of existing word embedding models requires a rather strong assumption of uniform word distribution. In this paper, we relax that assumption and propose more realistic conditions for proving additive compositionality, and we develop a novel word and sub-word embedding model that satisfies additive compositionality under those conditions. We then empirically show our model’s improved semantic representation performance on word similarity and noisy sentence similarity.