Large Language Models (LLMs) encode parametric knowledge but can exhibit format bias when resolving knowledge conflicts. When evidence is rendered as text, tables, or knowledge graphs, LLMs can favor certain formats despite unchanged semantic content. Existing work mainly diagnoses this bias, while existing inference-time mitigation remains tied to a single rendered prompt. We propose FaDe, a cross-format knowledge distillation framework that treats evidence format as a controllable nuisance variable and marginalizes teacher predictions over controlled format assignments. FaDe constructs a format-marginalized soft target from these renderings, then trains the student with an asymmetric two-path objective that combines heterogeneous-format debiasing with all-text anchoring. Experiments on ConflictBank and two QA-derived controlled conflict sets with four model backbones show FaDe reduces format bias while improving dual-source coverage and cross-format stability under single-pass inference. Our code is available at https://github.com/Kaia91/FaDe.

Large Reasoning Models (LRMs) achieve strong performance on table reasoning tasks but incur substantial inference cost due to long reasoning traces. Stepwise model routing mitigates this issue by dynamically assigning reasoning steps to smaller or larger models. However, stepwise model routing for table reasoning remains underexplored. Through empirical analysis, we find that reasoning steps involving tables contain two types of tokens with distinct uncertainty distributions: table tokens grounded in table structure, such as cell values and headers, and text tokens representing surrounding natural-language reasoning. The uncertainty of both token types is correlated with the risk that the model makes an error in the next reasoning step. However, existing methods fail to model them separately, leading to suboptimal routing decisions. To address this, we propose EcoTab, a table-aware stepwise routing framework for efficient table reasoning. For each current reasoning step, EcoTab separately models the two types of tokens and estimates their uncertainties, then combines them to estimate the risk of using the small model for the next step, thereby enabling model selection. Experiments on multiple table reasoning benchmarks show that EcoTab consistently outperforms strong baselines and achieves a better balance between accuracy and efficiency.

Stepwise model routing improves the efficiency of Large Reasoning Models (LRMs) by assigning each reasoning step to a suitable model. Recent methods formulate routing as a sequential decision process and train the router with reinforcement learning. However, although they model routing as a process, they still supervise the router with outcome rewards. Such rewards only reflect final answer correctness and fail to evaluate intermediate routing decisions, which can weaken performance and generalization. To address this gap, we propose RoRo, a rubric-guided process reward framework for stepwise model routing. RoRo first collects diverse routing trajectories and constructs preference pairs based on outcome, cost, and process quality. It then trains a Rubricor to generate a query-specific evaluation rubric and a Judge to score routing trajectories under this rubric through alternating optimization. The resulting process rewards are combined with outcome rewards to optimize the routing policy via GRPO. Experiments on five reasoning benchmarks under both same-family and cross-family settings show that RoRo consistently outperforms strong baselines and achieves better accuracy and cost trade-offs.