How Language Models Process Negation

We study how Large Language Models (LLMs) process negation mechanistically. First, we establish that even though open-weight models often provide wrong answers to questions involving negation, they do possess internal components that process negation correctly. Their poor accuracy is due to late-layer attention behavior that promotes simple shortcuts; ablating those attention modules greatly improves accuracy on negation-related questions. Second, we uncover how models process negation. We consider two hypotheses: models could use attention heads that attend to the phrase being negated and suppress related concepts, or they could directly construct a representation of the entire negative phrase (e.g., representing "not gas" as a vector that promotes liquids and solids). We apply a range of observational and causal interpretability techniques on Mistral-7B and Llama-3.1-8B to show that models implement both mechanisms, with the "constructive" mechanism being more prominent. Combined, our work deepens the understanding of LLMs' internals, highlighting construction-dominant computations and the coexistence of competing mechanisms within LLMs.

Convergent Evolution: How Different Language Models Learn Similar Number Representations

Language models trained on natural text learn to represent numbers using periodic features with dominant periods at $T=2, 5, 10$. In this paper, we identify a two-tiered hierarchy of these features: while Transformers, Linear RNNs, LSTMs, and classical word embeddings trained in different ways all learn features that have period-$T$ spikes in the Fourier domain, only some learn geometrically separable features that can be used to linearly classify a number mod-$T$. To explain this incongruity, we prove that Fourier domain sparsity is necessary but not sufficient for mod-$T$ geometric separability. Empirically, we investigate when model training yields geometrically separable features, finding that the data, architecture, optimizer, and tokenizer all play key roles. In particular, we identify two different routes through which models can acquire geometrically separable features: they can learn them from complementary co-occurrence signals in general language data, including text-number co-occurrence and cross-number interaction, or from multi-token (but not single-token) addition problems. Overall, our results highlight the phenomenon of convergent evolution in feature learning: A diverse range of models learn similar features from different training signals.

PDDL-Mind: Large Language Models are Capable on Belief Reasoning with Reliable State Tracking

Large language models (LLMs) perform substantially below human level on existing theory-of-mind (ToM) benchmarks, even when augmented with chain-of-thought prompting or probabilistic belief updates. We argue that these failures primarily arise from unreliable implicit state tracking rather than limitations in high-level reasoning. We introduce PDDL-Mind, a neuro-symbolic framework that decouples environment state evolution from belief inference. By translating narrative descriptions into explicit states and actions expressed in Planning Domain Definition Language (PDDL), and by verifying action-induced state transitions against a predefined domain, PDDL-Mind provides LLMs with a logically consistent and explicit representation of world states for ToM tasks. Experiments on MMToM-QA, MuMA and FanToM show that PDDL-Mind achieves over 5% absolute accuracy gain over the best existing state-of-the-art method on ToM benchmark questions.

Psychological Steering of Large Language Models

Large language models (LLMs) emulate a consistent human-like behavior that can be shaped through activation-level interventions. This paradigm is converging on additive residual-stream injections, which rely on injection-strength sweeps to approximate optimal intervention settings. However, existing methods restrict the search space and sweep in uncalibrated activation-space units, potentially missing optimal intervention conditions. Thus, we introduce a psychological steering framework that performs unbounded, fluency-constrained sweeps in semantically calibrated units. Our method derives and calibrates residual-stream injections using psychological artifacts, and we use the IPIP-NEO-120, which measures the OCEAN personality model, to compare six injection methods. We find that mean-difference (MD) injections outperform Personality Prompting (P$^2$), an established baseline for OCEAN steering, in open-ended generation in 11 of 14 LLMs, with gains of 3.6\% to 16.4\%, overturning prior reports favoring prompting and positioning representation engineering as a new frontier in open-ended psychological steering. Further, we find that a hybrid of P$^2$ and MD injections outperforms both methods in 13 of 14 LLMs, with gains over P$^2$ ranging from 5.6\% to 21.9\% and from 3.3\% to 26.7\% over MD injections. Finally, we show that MD injections align with the Linear Representation Hypothesis and provide reliable, approximately linear control knobs for psychological steering. Nevertheless, they also induce OCEAN trait covariance patterns that depart from the Big Two model, suggesting a gap between learned representations and human psychology.

Self-Evolving LLM Memory Extraction Across Heterogeneous Tasks

As LLM-based assistants become persistent and personalized, they must extract and retain useful information from past conversations as memory. However, the types of information worth remembering vary considerably across tasks. We formalize the \textit{heterogeneous memory extraction} task and introduce \textbf{BEHEMOTH}, a benchmark that repurposes 18 existing datasets spanning personalization, problem-solving, and agentic tasks, using a downstream utility-driven metric for systematic evaluation. Our empirical analysis confirms that no single static extraction prompt dominates across all task categories, and that existing self-evolving prompt optimization frameworks, originally designed for homogeneous distributions, degrade when training tasks are heterogeneous. To address this, we propose \textbf{CluE}, a cluster-based self-evolving strategy that groups training examples into clusters by extraction scenarios, analyzes each cluster independently, and synthesizes cross-cluster insights to update the extraction prompt. Experiments on BEHEMOTH show that CluE generalizes effectively across heterogeneous tasks ($+$9.04\% relative gain), consistently outperforming prior self-evolving frameworks.

ContextLeak: Auditing Leakage in Private In-Context Learning Methods

In-Context Learning (ICL) has become a standard technique for adapting Large Language Models (LLMs) to specialized tasks by supplying task-specific exemplars within the prompt. However, when these exemplars contain sensitive information, reliable privacy-preserving mechanisms are essential to prevent unintended leakage through model outputs. Many privacy-preserving methods are proposed to protect the information leakage in the context, but there are less efforts on how to audit those methods. We introduce ContextLeak, the first framework to empirically measure the worst-case information leakage in ICL. ContextLeak uses canary insertion, embedding uniquely identifiable tokens in exemplars and crafting targeted queries to detect their presence. We apply ContextLeak across a range of private ICL techniques, both heuristic such as prompt-based defenses and those with theoretical guarantees such as Embedding Space Aggregation and Report Noisy Max. We find that ContextLeak tightly correlates with the theoretical privacy budget ($ε$) and reliably detects leakage. Our results further reveal that existing methods often strike poor privacy-utility trade-offs, either leaking sensitive information or severely degrading performance.

Hubble: a Model Suite to Advance the Study of LLM Memorization

We present Hubble, a suite of fully open-source large language models (LLMs) for the scientific study of LLM memorization. Hubble models come in standard and perturbed variants: standard models are pretrained on a large English corpus, and perturbed models are trained in the same way but with controlled insertion of text (e.g., book passages, biographies, and test sets) designed to emulate key memorization risks. Our core release includes 8 models -- standard and perturbed models with 1B or 8B parameters, pretrained on 100B or 500B tokens -- establishing that memorization risks are determined by the frequency of sensitive data relative to size of the training corpus (i.e., a password appearing once in a smaller corpus is memorized better than the same password in a larger corpus). Our release also includes 6 perturbed models with text inserted at different pretraining phases, showing that sensitive data without continued exposure can be forgotten. These findings suggest two best practices for addressing memorization risks: to dilute sensitive data by increasing the size of the training corpus, and to order sensitive data to appear earlier in training. Beyond these general empirical findings, Hubble enables a broad range of memorization research; for example, analyzing the biographies reveals how readily different types of private information are memorized. We also demonstrate that the randomized insertions in Hubble make it an ideal testbed for membership inference and machine unlearning, and invite the community to further explore, benchmark, and build upon our work.

Zebra-CoT: A Dataset for Interleaved Vision Language Reasoning

Humans often use visual aids, for example diagrams or sketches, when solving complex problems. Training multimodal models to do the same, known as Visual Chain of Thought (Visual CoT), is challenging due to: (1) poor off-the-shelf visual CoT performance, which hinders reinforcement learning, and (2) the lack of high-quality visual CoT training data. We introduce $\textbf{Zebra-CoT}$, a diverse large-scale dataset with 182,384 samples, containing logically coherent interleaved text-image reasoning traces. We focus on four categories of tasks where sketching or visual reasoning is especially natural, spanning scientific questions such as geometry, physics, and algorithms; 2D visual reasoning tasks like visual search and jigsaw puzzles; 3D reasoning tasks including 3D multi-hop inference, embodied and robot planning; visual logic problems and strategic games like chess. Fine-tuning the Anole-7B model on the Zebra-CoT training corpus results in an improvement of +12% in our test-set accuracy and yields up to +13% performance gain on standard VLM benchmark evaluations. Fine-tuning Bagel-7B yields a model that generates high-quality interleaved visual reasoning chains, underscoring Zebra-CoT's effectiveness for developing multimodal reasoning abilities. We open-source our dataset and models to support development and evaluation of visual CoT.

When Do LLMs Admit Their Mistakes? Understanding the Role of Model Belief in Retraction

Can large language models (LLMs) admit their mistakes when they should know better? In this work, we define the behavior of acknowledging errors in previously generated answers as "retraction" and aim to understand when and why LLMs choose to retract. We first construct model-specific datasets to evaluate whether a model will retract an incorrect answer that contradicts its own parametric knowledge. While LLMs are capable of retraction, they do so only infrequently. We demonstrate that retraction is closely tied to previously identified indicators of models' internal belief: models fail to retract wrong answers that they "believe" to be factually correct. Steering experiments further demonstrate that internal belief causally influences model retraction. In particular, when the model does not believe its answer, this not only encourages the model to attempt to verify the answer, but also alters attention behavior during self-verification. Finally, we demonstrate that simple supervised fine-tuning significantly improves retraction performance by helping the model learn more accurate internal beliefs. Code and datasets are available on https://github.com/ayyyq/llm-retraction.

Textual Steering Vectors Can Improve Visual Understanding in Multimodal Large Language Models

Steering methods have emerged as effective and targeted tools for guiding large language models' (LLMs) behavior without modifying their parameters. Multimodal large language models (MLLMs), however, do not currently enjoy the same suite of techniques, due in part to their recency and architectural diversity. Inspired by this gap, we investigate whether MLLMs can be steered using vectors derived from their text-only LLM backbone, via sparse autoencoders (SAEs), mean shift, and linear probing. We find that text-derived steering consistently enhances multimodal accuracy across diverse MLLM architectures and visual tasks. In particular, mean shift boosts spatial relationship accuracy on CV-Bench by up to +7.3% and counting accuracy by up to +3.3%, outperforming prompting and exhibiting strong generalization to out-of-distribution datasets. These results highlight textual steering vectors as a powerful, efficient mechanism for enhancing grounding in MLLMs with minimal additional data collection and computational overhead.