The growing scale of evaluation tasks has led to the widespread adoption of automated evaluation using LLMs, a paradigm known as "LLM-as-a-judge". However, improving its alignment with human preferences without complex prompts or fine-tuning remains challenging. Previous studies mainly optimize based on shallow outputs, overlooking rich cross-layer representations. In this work, motivated by preliminary findings that middle-to-upper layers encode semantically and task-relevant representations that are often more aligned with human judgments than the final layer, we propose LAGER, a post-hoc, plug-and-play framework for improving the alignment of LLM-as-a-Judge point-wise evaluations with human scores by leveraging internal representations. LAGER produces fine-grained judgment scores by aggregating cross-layer score-token logits and computing the expected score from a softmax-based distribution, while keeping the LLM backbone frozen and ensuring no impact on the inference process.

Our experimental evaluations demonstrate that this simple modification significantly improves the quality of localization maps on both the P ASCAL VOC 2012 and MS COCO 2014 datasets, exhibiting a new state-of-the-art performance for weakly supervised semantic segmentation.

These molecular properties were calculated using a hybrid quantum simulation (Gaussian 09) at the B3LYP/6-31G(2df,p) level of theory. In this study, we created a subset of the QM9 dataset with a limited number of atoms, M 14, per molecule, which we refer to as the "QM9under14atoms" dataset in the main text. As the learning/predicting targets, we selected three kinds of energy properties: atomization energy at 0 K, zero point vibrational energy, and enthalpy at 298.15 K. E RN is the bias vector in layer ℓ. The LCAO considers the normalization for the coefficients in Eq. (6) in the main text. Additionally, the normalization term in Eq. (7) in the main text is calculated as follows: Z(qn,ζn)=

Protein language models (PLMs) have transformed sequence-based protein analysis, yet most applications rely only on final-layer embeddings, which may overlook biologically meaningful information encoded in earlier layers. We systematically evaluate all 33 layers of ESM-2 for kinase functional prediction using both unsupervised clustering and supervised classification. We show that mid-to-late transformer layers (layers 20-33) outperform the final layer by 32 percent in unsupervised Adjusted Rand Index and improve homology-aware supervised accuracy to 75.7 percent. Domain-level extraction, calibrated probability estimates, and a reproducible benchmarking pipeline further strengthen reliability. Our results demonstrate that transformer depth contains functionally distinct biological signals and that principled layer selection significantly improves kinase function prediction.

Do neural networks build their representations through smooth, gradual refinement, or via more complex computational processes? We investigate this by extending the logit lens to analyze the policy network of Leela Chess Zero, a superhuman chess engine. Although playing strength and puzzle-solving ability improve consistently across layers, capability progression occurs in distinct computational phases with move preferences undergoing continuous reevaluation--move rankings remain poorly correlated with final outputs until late, and correct puzzle solutions found in middle layers are sometimes overridden. This late-layer reversal is accompanied by concept preference analyses showing final layers prioritize safety over aggression, suggesting a mechanism by which heuristic priors can override tactical solutions.