arXiv:2609. 23366v1 Announce Type: new Abstract: Recurrent models must preserve information that changes future behavior while suppressing hidden-state error.
arXiv:2609.06036v1 Announce Type: new Abstract: Proposal-based controllers---learned policies, language-model planners, and other black-box \emph{generators}---are increasingly deployed behind runtim...
arXiv:2607. 18305v1 Announce Type: cross Abstract: Some limits on what language models know are not gaps in data coverage but structural properties of learning from text.
The paper investigates whether a small, directly addressable change in the hidden state of a learned world model can steer its future predictions along a desired counterfactual trajectory. Using a 192‑dimensional recurrent model in a two‑object collision setting, the authors identify low‑rank latent carriers—specifically a rank‑4 patch—that, when applied, successfully redirect a 12‑step autonomous rollout without further intervention. The study demonstrates that this compact intervention interface consistently works across independently trained checkpoints and intervention times, while various control experiments confirm the specificity of the effect.
arXiv:2609.37841v1 Announce Type: new Abstract: Masked generative models offer parallel token prediction, but accurate parallel sampling must account for dependencies among tokens. When dependencies...
arXiv:2606. 25449v1 Announce Type: cross Abstract: A language model's memory can be worse than having no memory at all.
BARRIER (Bounded Activation Regions for Robust Information Erasure) is a method for machine unlearning that confines parameter updates to a controlled activation space, allowing stronger erasure of targeted concepts while limiting collateral damage to other representations. By employing interval arithmetic, it derives a closed‑form bound on worst‑case representation changes in protected regions, which serves as a knowledge‑preservation objective. The approach is architecture‑agnostic, compatible with existing erasure objectives, and empirically shows competitive performance in both classification and generative tasks, with enhanced robustness against adversarial recovery attacks.
arXiv:2512. 23075v5 Announce Type: replace-cross Abstract: Policy gradient methods for Large Language Models optimize a policy $\pi_\theta$ via a surrogate objective computed from samples of a rollout policy $\pi_{\text{roll}}$.
The paper presents an interactive probabilistically checkable proof (PCP) protocol that allows a polynomial‑time verifier to check the approximate consistency of a probabilistic predictor defined by two circuits, P and Q. By evaluating these circuits at a few points and querying a proof oracle that encodes a witnessing probability distribution, the verifier can confirm that the predictor’s many conditional‑probability claims are self‑consistent. The authors also establish that the problem of verifying l₂‑approximate consistency for explicit probabilistic claims lies in NP, with certificates of size O(mn + log B), and show how to eliminate dependence on the input bit‑precision B through a small additive gap.
arXiv:2609.07065v1 Announce Type: new Abstract: Agent-to-agent (A2A) alpha discovery is slowed by repeated feedback cycles between mining and evaluation agents, whose hand-offs, in contemporary LLM m...
The paper investigates the safety of Code World Models, where a language model generates an executable world model that a planner uses, and the model is accepted if it reproduces sampled transitions. It defines the pipeline’s danger as the expected risk, showing that the probability of missing a critical event across N independent rollouts is (1‑r)^N, and that an additional acceptance sample adds to the exponent. Experiments on hybrid instruments reveal that mode‑blind models can be exploited, and the authors provide theoretical bounds on localization budgets and demonstrate that acceptance only guarantees sample consistency, covering about two percent of the planner’s queries.
arXiv:2608.23034v1 Announce Type: cross Abstract: Controlling restricted knowledge in large language models is essential for model alignment and safe deployment. Test-time unlearning avoids costly re...
A language model's memory can be worse than having no memory at all. Give a model a memory that kept a wrong conclusion but dropped the work behind it, and it emits that stale value as a confident answer; give the same model an empty memory and it abstains.