4arXiv cs.CL (Computation and Language)·
Study evaluates LLM and human interpretation of coded language in cybercrime Discord chats
Researchers present an exploratory study on interpretation difficulty in cybercrime-related Discord chats, focusing on slang, coded terms, and community-specific expressions. The study constructs expert-reviewed reference interpretations and uses them to benchmark both human and LLM performance under varying context conditions. Key findings: local context alone is insufficient for humans, while extended conversational context and external knowledge substantially help; larger LLMs outperform smaller ones; the authors propose a preliminary taxonomy of factors driving interpretation difficulty. The work argues that harmful-content analysis should be framed as evidence-integration rather than message-level classification.
Related guides (2)
Related events (8)
LLMs susceptible to cognitive heuristics in code vulnerability detection, enabling black-box attacks
A new arXiv paper presents the first systematic study of cognitive heuristics — halo effect, framing effect, and anchoring — in LLM-based code vulnerability detection. Evaluating eight LLMs across three programming languages, the authors find all models susceptible, with framing causing the largest average shift (33.2%), followed by anchoring (23.5%) and halo (18.4%). Critically, the paper demonstrates a proof-of-concept black-box attack that suppresses up to 97% of previously detected vulnerabilities by manipulating surrounding context without changing the code itself. The findings suggest cognitive susceptibility is a systematic and exploitable property of LLM security tooling.
Paper challenges LLM expert-level claims by measuring variance and error magnitude in code-based data analysis tasks
A new arXiv paper argues that standard LLM benchmarks overstate model capabilities by focusing on average performance on training-data-adjacent tasks while ignoring response variance and error magnitude. The authors introduce a novel benchmark requiring frontier LLMs to write code for data analysis tasks, comparing results against human expert submissions. Human experts outperformed the frontier LLM on average across multiple metrics and showed lower performance variability. The findings challenge the prevailing narrative that LLMs perform at human-expert level on knowledge economy tasks.
LLMs fail to consistently simulate demographic perspective-taking in hate speech annotation
A new arXiv paper evaluates whether persona-conditioned LLMs can replicate how different demographic groups perceive hate speech, testing three dimensions: inter-group disagreement, in-group sensitivity, and vicarious prediction. No model consistently captures all three dimensions, and performance is highly model-dependent rather than emerging reliably from identity prompts alone. Vicarious prompting with Llama 3.1 provides the closest approximation to human disagreement patterns across demographic axes. The findings have implications for using LLMs as proxies for diverse human annotators in content moderation tasks.
Mechanism-oriented taxonomy of indirect linguistic encoding improves LLM-based coded language detection
Researchers propose a comprehensive taxonomy of indirect linguistic expressions (ILE) — covering algospeak, euphemisms, and adversarial obfuscation — organized by underlying encoding mechanisms rather than communicative intent. The taxonomy is evaluated by injecting it into LLM prompts and benchmarked against four existing taxonomies and a no-taxonomy baseline on 2,000 manually annotated TikTok and Bluesky posts. The proposed taxonomy achieves 4.7% accuracy and 5.4% F1 improvements over the best competing approach across three LLMs, suggesting structured linguistic scaffolding meaningfully aids content moderation tasks.
Survey chapter on LLM mechanisms, emergent capabilities, and cognition debates
A new arXiv preprint surveys current understanding of large language models, covering the Transformer architecture, emergent capabilities resembling human cognition (symbolic reasoning, theory of mind, deception), and explainability approaches from neuron activation analysis to circuit tracing. The chapter also engages the debate over whether LLMs genuinely understand or merely pattern-match, arguing against reductive anti-anthropomorphism while acknowledging human-LLM differences. It is framed as a book chapter synthesizing recent empirical findings and theoretical positions.
Study of security and privacy prompts in the wild reveals LLM response quality gaps and inconsistency
Researchers analyzed 14,727 security and privacy (S&P) prompts drawn from WildChat's 3.2M real user-LLM conversations, categorizing them into nine topic areas and evaluating response quality across 270 advice-seeking prompts. Commercial models substantially outperformed open-weight models (GPT achieving 98% 'good enough' responses vs. Llama 4 at 47%), but even high-performing commercial models showed inconsistent responses across repeated runs of the same prompt. The study is the first to analyze real user S&P queries to LLMs rather than expert-authored test sets, surfacing both a capability gap and a reliability concern.
LLMs Show Inverted Compositional Strengths vs. Humans on Reference Resolution Task
This paper evaluates LLMs and humans on the Personal Relation Task (Paperno 2022), distinguishing between Extensional tasks (resolving what an expression refers to) and Intensional tasks (representing structured sense/formula). The study finds that humans outperform LLMs on Extensional tasks while LLMs outperform humans on Intensional tasks—an inverted pattern of strengths. The authors argue this asymmetry reflects the absence of referential grounding in LLM training as a key gap in human-like language understanding.
Tracing the Emergence of Human-Like Pragmatic Reasoning in LLMs Across Languages
Researchers conducted a population-matching experiment evaluating 25 LLMs on conditional inference tasks across four languages, comparing model behavior to matched human populations. The study finds that LLMs function as accurate semantic operators but systematically fail to capture pragmatic enrichments—context-sensitive inferences beyond literal logical meaning—that humans apply effortlessly. Model performance on pragmatic reasoning is not predicted by open vs. closed weights, training orientation, or architecture type, suggesting pragmatic reasoning remains an emergent and unreliable capability. The findings contribute to ongoing debates about whether LLMs reason like humans or merely approximate surface-level linguistic patterns.