Reasoning-conditioned embeddings can improve retrieval by progressively refining semantic constraints. However, additional reasoning may introduce over-generalized or semantically noisy abstractions that weaken retrieval specificity. Existing methods largely supervise reasoning through final retrieval outcomes, providing limited guidance over the refinement process itself. We propose AURA, a framework that directly supervises intermediate reasoning states according to how well they separate relevant documents from hard distractors. AURA further trains a lightweight utility estimator over prefix-conditioned embeddings to adapt reasoning depth at inference time. On the reasoning-intensive BRIGHT benchmark, AURA-4B achieves 34.7 average nDCG@10 while using 2.8 reasoning steps and 96 generated tokens per query. Under matched deployment conditions, adaptive stopping reduces latency from 4.6 to 0.76 seconds relative to full-trajectory reasoning. Results on MS MARCO further suggest that this specialization need not substantially degrade conventional in-domain retrieval.

Automated LLM workflow design often relies on expensive per-task search, which is computationally expensive and does not explicitly reuse structural knowledge across tasks. We propose SWIFT (Synthesizing Workflows via Few-shot Transfer), which reuses prior search trajectories and demonstrations to synthesize target-task workflows without iterative target-task optimization. SWIFT distills compositional heuristics and output-interface contracts from source-task trajectories, then conditions a single LLM generation pass on these priors and cross-task workflow demonstrations to produce an executable workflow. Across 5 benchmarks, SWIFT outperforms the SOTA search-based baseline while reducing marginal per-task optimization cost by three orders of magnitude given an existing source bank. It also improves performance on 4 additional held-out benchmarks and transfers from GPT-4o-mini to Grok, Qwen, and Gemma execution backbones. These results suggest that prior workflow search can expose reusable structural priors that reduce optimization cost while preserving or improving accuracy on related tasks.

As large language models (LLMs) are increasingly deployed in real-world applications, alignment is no longer governed by a single universal notion of safety or helpfulness, but instead by provider- or application-specific model specifications. These specifications are typically long, structured, and frequently updated, yet existing alignment pipelines lack a systematic mechanism to operationalize them as training signals. In this paper, we propose specification-grounded alignment, a new alignment paradigm that treats provider-authored model specifications as the primary alignment target rather than abstract principles or static benchmarks. To instantiate this paradigm, we introduce SpecAlign, a framework that synthesizes alignment data directly from specification documents. SpecAlign combines structured rule annotation, controllable specification instantiation, and multi-agent adversarial data synthesis to generate fine-grained, boundary-aware preference pairs that capture both compliant behaviors and meaningful specification violations. Experiments across multiple model specifications and backbone models demonstrate that training with SpecAlign consistently improves rule compliance while preserving general capabilities and avoiding over-conservative behavior. These results suggest that grounding alignment in explicit model specifications enables rapid, precise, and scalable adaptation of LLM behavior to evolving policy requirements.