arXiv:2608. 20054v1 Announce Type: new Abstract: Multi-module systems often expose every module to the full input.
arXiv:2608. 08829v1 Announce Type: cross Abstract: Activation steering edits the behaviour of a frozen language model by adding a learned vector to its residual stream, and current practice fixes the injection layers globally per task.
arXiv:2608. 17564v1 Announce Type: cross Abstract: Unified multimodal models (UMMs) are motivated by the hope that understanding and generation reinforce each other but controlled ablations repeatedly find that adding a generation objective leaves understanding flat.
arXiv:2608. 05162v1 Announce Type: cross Abstract: Pooling is a consequential but under-examined design choice in decoder-only concept representation work: practitioners must collapse token-level hidden states into a passage-level vector, yet no shared protocol exists for comparing this choice across concepts, models, and tasks.
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.
The paper introduces Language Identity Head Ablation (LIHA), a causal method that zeroes individual attention heads in transformer models to measure language switch rates across multilingual prompts. Applying LIHA to GPT‑2 reveals a small set of first‑token broadcaster heads—most notably L6H1—that persistently attend to the initial prompt token and propagate language signals throughout generation, with compensatory head recruitment occurring hierarchically in higher layers. A controlled comparison between Qwen2.5‑1.5B‑Base and Qwen2.5‑1.5B‑Instruct shows that instruction tuning concentrates language‑identity influence in early layers, while experiments with Chinese and Russian confirm script‑specific first‑token broadcasting at layer 0.
arXiv:2608. 04618v1 Announce Type: new Abstract: Heterogeneous language-model ensembles expand the space of candidate responses, yet they lack a principled criterion for when a newly generated answer should supersede an already supported one.
arXiv:2603. 18482v2 Announce Type: replace-cross Abstract: Standard decoding strategies for text generation, including top-$k$, nucleus sampling, and contrastive search, select tokens based on likelihood, restricting outputs to high-probability regions.
arXiv:2607. 20436v1 Announce Type: cross Abstract: Safety evaluations often assume that behavior observed during testing reflects behavior in ordinary use, but fine-tuning can break this assumption.
arXiv:2606. 21876v2 Announce Type: replace-cross Abstract: The Categorical Jacobian of Zhang et al.
arXiv:2303. 15619v2 Announce Type: replace-cross Abstract: The choice of \emph{which} tokens to mask is a central, under-examined design decision in masked language modeling (MLM).
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.
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.