arXiv:2608. 15809v1 Announce Type: new Abstract: Behavioral accuracy, linear decodability, and successful activation interventions do not by themselves show that a model carries an operation-level structure from one symbolic domain to another.
arXiv:2608. 13626v1 Announce Type: new Abstract: A hidden state signal can be decodable or causally usable without supporting a reusable action map.
arXiv:2608. 03842v1 Announce Type: cross Abstract: When a language model fails on surface-perturbed input (typos, OCR noise, homophones), "which layer is responsible" has three natural operationalizations: where representations diverge most (sensitivity), where restoring clean activations recovers the prediction (causality), and where a small adapter can repair the damage (compensatory capacity) - and we show these three layer maps dissociate.
arXiv:2605. 07284v2 Announce Type: replace Abstract: A late-layer change learned during post-training may work on the base model's earlier state, or it may depend on earlier computation learned with it.
arXiv:2608. 11822v1 Announce Type: cross Abstract: A growing body of work reports that language models represent task-relevant latent structure that they fail to use.
arXiv:2608. 11797v1 Announce Type: new Abstract: Model merging by task arithmetic works until it doesn't, and the field diagnoses why with magnitudes: layerwise representation bias, deviations from cross-task linearity, parameter overlap.
arXiv:2607. 29484v1 Announce Type: cross Abstract: Interventional data is widely regarded as the gold standard for teaching models causal reasoning.
arXiv:2607. 04510v1 Announce Type: cross Abstract: Emergent misalignment (EM) -- the broad misbehaviour a language model acquires after fine-tuning on narrow harmful data -- is mediated in Qwen2.
arXiv:2606. 05378v1 Announce Type: new Abstract: We test whether a single screen-and-ablate recipe -- identify attention-head circuits by task-pattern selectivity, then verify by causal ablation against a matched-random null -- produces consistent mechanistic claims across model families.
arXiv:2605. 09692v3 Announce Type: replace Abstract: Autonomous language agents increasingly expose traces, memories, plans and constraints, but existing evaluations rarely test whether these state variables are bound to final actions.
arXiv:2608. 09490v1 Announce Type: new Abstract: Task arithmetic treats fine-tuning displacements as composable directions in weight space, yet it remains unclear when parameter addition reflects predictable changes in model function.
arXiv:2606. 08292v2 Announce Type: replace Abstract: Mechanistic studies often assign a component a role when removing it damages a behavior, its activation linearly encodes task information, and restoring that activation repairs the damage.
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.