arXiv:2609.25602v1 Announce Type: new Abstract: In language models, the choice between believing the prompt and believing the weights is made by a handful of identifiable attention heads. Instruction...
The paper investigates how large language models balance instruction-following with pattern completion when the two objectives conflict. By creating dialogues where a user instruction to act in a target way T is opposed by assistant turns that demonstrate a competing pattern P, the authors measure instruction-following rates across 13 models and 16 instructions over up to 50 turns. Results show wide variability (1%–99%) in instruction adherence, with robustness influenced by instruction content, output format, and chain-of-thought reasoning, but overall instruction-following remains brittle under induction pressure.
arXiv:2606. 15420v1 Announce Type: cross Abstract: A constitution tells a language model what to value, but little tells us whether it does.
arXiv:2606. 07808v1 Announce Type: new Abstract: Reasoning language models deployed in agentic workflows must follow an instruction hierarchy: when instructions from different sources conflict, the model should obey the highest-privilege applicable instruction.
arXiv:2608. 13921v1 Announce Type: new Abstract: LLM agents increasingly maintain personal memory across sessions, but it can conflict.
arXiv:2607. 24765v1 Announce Type: cross Abstract: Large language models (LLMs) can give different answers to the same decision problem across runs, and reverse a decision when their own prior answer returns as context.
LLM agents increasingly maintain personal memory across sessions, but it can conflict. Preferences depend on context, behavior evolves, and sources can conflict. When a query lacks context, time, or s...
The paper treats human interventions during LLM agent failures as labels that pinpoint the internal computation that went wrong, rather than merely training data. In a controlled environment, the authors analyze one failure type—over‑asking a clarifying question when the request is already clear—using rule‑based overseer labels validated by blind human judgment. They find that the decision to ask versus act can be causally localized across several models, that over‑asking exhibits distinct mechanistic signatures, and that steering or weight consolidation can mitigate the issue, though outcomes vary by model family.
The paper presents a defense‑in‑depth framework for large language models that separates internal activation steering from external memory handling to combat sycophancy. It evaluates four open‑weight models on a new MemSyco‑Bench dataset, testing five memory‑defense configurations—including a Router Gate that selectively rewrites, keeps, or drops memories—and measures sycophancy and accuracy across 1,550 items. Results show that selective Router Gate filtering preserves more accuracy than complete memory removal, while inverse steering slightly reduces sycophancy but is not statistically significant.
The paper introduces ELCD, a latent conflict detector that verifies LLM outputs after generation to catch instruction conflicts that static input checks miss. ELCD builds a hidden-state representation from the final-token embedding and the mean-pooled response embedding, then trains a pairwise margin ranking objective to distinguish compliant from drifting responses. Experiments on five large language models show ELCD outperforms baselines, boosting PR-AUC for Llama‑2‑7B by ~30 percentage points and cutting FPR95 for Mistral‑7B to 2.67%.
The paper introduces Many-Tier Instruction Hierarchy (ManyIH), a new framework for resolving conflicts among instructions with arbitrarily many privilege levels in large language model agents. It presents ManyIH-Bench, a benchmark featuring 853 agentic tasks that require navigating up to 12 levels of conflicting instructions across 46 real-world agents. Experiments show current models achieve only about 40% accuracy when instruction conflict scales, highlighting a gap in fine-grained, scalable conflict resolution.
arXiv:2609.25686v1 Announce Type: cross Abstract: Long-horizon assigned work requires an LLM agent to track the state of a task: which steps are done, blocked, cancelled, or open to repetition. Agent...