arXiv:2606. 12923v1 Announce Type: cross Abstract: AI alignment, interpretability, steering, and neural perturbation studies identify order-inducing objects.
arXiv:2606. 12923v2 Announce Type: replace-cross Abstract: AI alignment, interpretability, steering, and neural perturbation studies identify order-inducing objects.
NeuronSifter is a framework for planning interventions in central nervous system microenvironments by converting treatment regimens into state‑conditional target‑occupancy fields and propagating them through microenvironment dynamics. It selects measurements based on their expected reduction in intervention loss, integrating typed outcomes into a unified posterior. In synthetic Alzheimer’s disease scenarios, occupancy conditioning improves trajectory probability scores and intervention ordering accuracy, and decision‑directed acquisition reduces terminal risk compared to a Bayesian experimental design planner.
arXiv:2608. 19338v1 Announce Type: cross Abstract: Mechanistic interpretability seeks quantities that models do not expose directly: represented states, component effects, interactions, and responses to interventions.
arXiv:2606. 19831v1 Announce Type: cross Abstract: Aligned language models gate behaviors such as refusal and language routing through sparse feed forward neurons, yet no theory predicts when a single neuron intervention controls a behavior coherently rather than collapsing the output.
ObserverBench is a benchmark framework that evaluates whether internal mechanistic estimators—called observers—are suitable for guiding interventions, control, or safety actions in language models. It separates estimation accuracy from the loss incurred by the chosen action, showing that accurate predictions do not always lead to better decisions. Experiments on GPT‑2‑small, Qwen2.5‑7B, Gemma‑2‑9B‑it, and Qwen3.5‑9B demonstrate that observers trained on action loss can reduce deployment loss, while traditional metrics like AUROC may rank monitors differently from actual performance.
arXiv:2607. 27849v1 Announce Type: cross Abstract: An open-weight LLM can write composition setpoints every five minutes.
arXiv:2608. 01548v2 Announce Type: replace Abstract: Language-first intelligence is constrained by which distinctions enter its symbolic record, which mappings its language--interpreter--environment complex can execute, and which possibilities can be realized with finite resources.
arXiv:2607. 09156v1 Announce Type: new Abstract: Additive activation steering (injecting a scaled residual-stream direction during generation) is calibrated almost entirely in single-turn chat, yet the models it targets are increasingly deployed as tool-using ReAct agents.
LM‑X is a generalist vision‑language‑action policy that augments action prediction with three online, explicitly supervised signals: return‑to‑go (RTG) for task progress, event‑to‑go (ETG) for the next semantic transition, and heteroscedastic action flow for local reliability. By conditioning action generation on these signals, LM‑X embeds explainability directly into control rather than as a post‑hoc explanation. After a 20‑day pretraining run on 64 GPUs, LM‑X outperforms an action‑only backbone by 16.0 points and a single‑head variant by 10.8 points, and achieves 74.1 % success on 50 RoboTwin2.0 tasks and 68.6 % on seven real‑robot tasks, surpassing the GR00T N1.7 baseline.
arXiv:2607. 24339v1 Announce Type: new Abstract: Large language model (LLM) agents inherit reactive failure modes: escalation under provocation, sycophantic drift under flattery, perseveration when stuck.
arXiv:2606. 19111v1 Announce Type: cross Abstract: Team science holds that leadership is contingent: it helps only under specific conditions, and capable, autonomous teams may need none at all.
ObserverBench is a benchmark framework that evaluates whether internal mechanistic estimators—called observers—are suitable for guiding interventions, control, or safety actions in language models. It separates estimation accuracy from the loss incurred by the chosen action, demonstrating that accurate average estimates can still lead to poor decisions. Experiments on GPT‑2‑small, Qwen2.5‑7B, Gemma‑2‑9B‑it, and Qwen3.5‑9B show that observers trained on action loss tend to select lower‑loss actions, while traditional metrics like AUROC can rank monitors differently from deployment loss, highlighting the need for task‑specific evaluation.