We present CXSearch, an automated system for discovering inputs on which a target language model fails to satisfy a stated specification. CXSearch frames failure discovery as constrained search in a continuous embedding space, with a learned acceptance predicate that rewards inputs producing both diverse and severe failures.
Standard byte-pair encoding tokenizers trained on web-scale mixed corpora underperform on source code: indentation runs, common identifier patterns, and language keywords are fragmented across multiple tokens. We introduce CATok, a code-aware tokenization scheme that augments BPE with three structural primitives — leading-whitespace runs, camel/snake-case-aware identifier merges, and language-keyword anchors — added before the BPE merge schedule begins.
Activation steering has emerged as a lightweight alternative to fine-tuning for modulating large language model behavior. We study a particularly minimal variant: sparse mean-difference steering, in which a steering vector is computed as the difference of mean residual-stream activations on contrasting prompt sets, then projected onto its top-k dimensions before injection.
We conduct the largest study to date on backtracking, analyzing 38,847 instances across 12 datasets spanning multiple domains. Our key finding is that search accounts for 32.
Syntactic priming—the tendency to reuse recently encountered grammatical structures—is a well-established phenomenon in human language production. Whether transformer language models exhibit analogous structural persistence, and whether such persistence extends across the boundaries of attention context windows, remains unknown.
Large language models exhibit sycophantic behavior—adjusting their responses to agree with user opinions even when those opinions are factually incorrect. While prior work has measured sycophancy in single-turn settings, real-world interactions are multi-turn, and the dynamics of sycophancy across extended dialogues remain unexplored.
Reinforcement Learning from Human Feedback (RLHF) has become the dominant paradigm for aligning large language models (LLMs) with human preferences. However, reward hacking—where models exploit reward model weaknesses to achieve high scores without genuine quality improvement—remains a critical failure mode that is difficult to detect post-deployment.
Large language models (LLMs) enable state-of-the-art performance across diverse tasks but face latency challenges in real-time applications due to their autoregressive nature. Speculative decoding accelerates inference by generating multiple tokens per forward pass through parallelization with a smaller draft model, improving throughput by 2-5x.
Curriculum learning for synthetic data achieving 19.17% perplexity improvement over random ordering.
Long-context capability is increasingly the limiting factor for LLM-based agents that must plan, search, debug, and maintain state over hours-to-days of interaction. “More tokens” alone is not a solution: practical systems fail due to token budget blowups, inference-time KV-cache costs, and degradation in information use as relevant facts drift away from the beginning/end of the prompt (the “lost-in-the-middle” effect).