arXiv:2605. 11554v2 Announce Type: replace Abstract: Task-agnostic structure proxies are often used to interpret why one pretraining corpus transfers better than another, but such explanations require the proxy to track the structure that matters for the downstream task.
arXiv:2609.39702v1 Announce Type: new Abstract: Adapting a language model to a specialized corpus means choosing which instruction-tuning tasks to train on under a fixed budget, and testing one choic...
arXiv:2608. 13069v1 Announce Type: new Abstract: Large language models (LLMs) are predominantly aligned to function as passive, sycophantic assistants.
Large language models (LLMs) are predominantly aligned to function as passive, sycophantic assistants. We challenge this default paradigm by empirically evaluating the cognitive plasticity of open-weight architectures when subjected to rigorous behavioral reprogramming.
The study investigates how layer‑wise intervention responses in language models change over the course of pretraining, using single‑block identity bypass across multiple checkpoints and model‑domain combinations. It finds that while depth ordering of responses persists, their magnitudes shift, with nearby checkpoints showing stronger rank correspondence than distant ones and large changes occurring at positions that recur across samples and transfer across evaluation domains. Controlled experiments reveal that these longitudinal changes cannot be explained by a single downstream sensitivity and depend on perturbation strength and direction, indicating that layer sensitivity is structured but dynamic.
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 study investigates how post‑training of large autoregressive language models (ARMs) into masked diffusion models (MDMs) affects their internal computation. Across two 7B ARM‑MDM families and four diagnostic tasks, the authors find that MDMs retain much of the ARM’s high‑attribution pathways on prefix‑dominant tasks, but reorganize computation toward earlier layers on globally constrained tasks. Component‑level probes reveal that ARMs depend on sharply specialized components, whereas MDMs show weaker specialization and more diffuse output‑space alignment.
Measuring training data influence consistently across language model pretraining is challenging. It is difficult to select downstream tasks or validation sets representative of a model's general capabilities, and reliance on task performance at intermediate checkpoints complicates comparisons across training.
arXiv:2608. 16347v1 Announce Type: cross Abstract: Direct communication between AI systems relies on natural language as an intermediate layer, incurring encoding/decoding overhead, token cost, and latency.
arXiv:2607. 21692v1 Announce Type: new Abstract: Sparse attention reduces the cost of long contexts by allowing each query to read only selected parts of the input.
The paper investigates how synthetic pretraining priors used in tabular foundation models (TFMs) influence downstream performance. By reconstructing the synthetic data generators of four TFMs and comparing their generated tasks to two popular tabular benchmarks using structural descriptors, the authors measure structural coverage and normalized density. They find that some generators provide broader and denser support for benchmark tasks, and that stronger synthetic-to-benchmark support generally correlates with better model performance.
TabCausal is a causal discovery foundation model that learns to map datasets directly to causal graphs by pretraining across diverse causal environments. It uses a dynamic task construction strategy to expose the model to varied graph priors, mechanisms, noise models, dimensions, sample sizes, and intervention regimes, improving transferability from observational and mixed‑interventional data. On large synthetic benchmarks and a new protocol‑guided semantic benchmark, TabCausal outperforms many classical baselines and shows robust structure recovery, especially when interventional evidence is available.
The paper "Stress-testing Alignment Midtraining" examines the effectiveness of alignment midtraining (AMT), a technique that continues pretraining on alignment-relevant data to improve generalisation beyond post‑training methods. Experiments on models up to 110 billion parameters and 1 billion midtraining tokens reveal that AMT can steer a model’s motivation in simple scenarios, but its effects are quickly overridden by even a tiny fraction of finetuning data with a competing motivation. The study also shows that rule-following requires demonstrations in either the midtraining or post‑training datasets to be robustly learned, leading the authors to conclude that current public evidence is insufficient to confirm that AMT resolves the core alignment challenges of powerful AI systems.