Large language model (LLM) pricing agents may respond to how market data is presented, even when its numerical values remain unchanged. We introduce market signal injection (MSI), an attack that manipulates numerical formatting, competitor ordering, or qualitative market commentary without issuing explicit instructions. We evaluate nine open-weight models in simulated Bertrand duopoly and triopoly markets and three proprietary models in duopoly markets. Sentiment-based attacks produce the largest behavioral shifts, which propagate to other firms and alter profits and consumer surplus. Susceptibility varies across model families, and larger models are not consistently more robust. Matched neutral-text controls and a rule-based agent support a framing-based account of these shifts under the fixed demand parameters of our simulation. Episode-held-out probes distinguish baseline from attacked activations in all eleven re-evaluated model–condition pairs: linear AUC is 1.00 and MLP AUC ranges from 0.93 to 0.99. This separability does not by itself identify harmful pricing decisions. Input canonicalization removes the tested sentiment attacks, while decision boundary anchoring, which combines prompt constraints with output projection, provides partial mitigation under the tested adaptive attacks. These results identify data presentation as an attack surface for LLM pricing agents and motivate defenses that account for interactions among agents.

Large language models (LLM) deployed as autonomous pricing agents may sustain supracompetitive prices through tacit coordination. We develop a causal graph divergence framework that separately measures structural faithfulness and intent faithfulness of LLM pricing agents in Bertrand competition. Across nine LLMs under duopoly and triopoly conditions, collusive behavior and chain-of-thought (CoT) faithfulness dissociate along both dimensions: the most collusive model accurately reports cooperative intent yet reasons structurally unfaithfully, while the most structurally faithful model sustains supra-Nash pricing under both market structures. These findings establish that CoT monitoring alone cannot serve as a standalone safeguard against algorithmic collusion.

Large Language Models (LLMs) have demonstrated remarkable capabilities across various domains, yet they struggle with causal reasoning tasks. We introduce PACKaGE (Paragraph-level Annotated Causal Knowledge and Graph–tExt), a causal benchmark dataset that pairs paragraph-level texts with explicit causal graphs. Unlike prior sentence-level resources, our benchmark dataset (i) operates at paragraph level to capture various and interdependent causal pathways; (ii) offers multimodal annotations including full graphs and contextualized relations; and (iii) maps extended passages to comprehensive graphs. We detail a four-stage construction pipeline: collecting causality-labeled and fact-centric texts, inducing graphs from span-tagged pairs, condensing nodes, and recomposing path-aligned narratives. Overall, the paired evaluation highlights a persistent discrepancy between linguistic competence and causal understanding in LLMs, pointing to the need for further development.