https://tolgadimli.github.io/Recent content on Tolga DimliogluHugo -- 0.147.2enTue, 18 Nov 2025 00:00:00 +0000https://tolgadimli.github.io/indcollab/data1/Tue, 18 Nov 2025 00:00:00 +0000https://tolgadimli.github.io/indcollab/data1/A modular retrieval-augmented LLM framework for automating contract revision analysis and optimization under low-resource, real-world industrial settings.https://tolgadimli.github.io/papers/paper1/Sat, 15 Nov 2025 00:00:00 +0000https://tolgadimli.github.io/papers/paper1/We propose a retrieval-augmented LLM system that automatically identifies and rewrites problematic industrial contract revisions, achieving high accuracy even in low-resource legal settings.https://tolgadimli.github.io/papers/paper2/Tue, 15 Jul 2025 00:00:00 +0000https://tolgadimli.github.io/papers/paper2/We propose a communication-efficient distributed training framework (DPPF) that introduces a pushing force to promote wide minima recovery, improving generalization without incurring significant computational or communication overhead.https://tolgadimli.github.io/papers/paper3/Wed, 15 Jan 2025 00:00:00 +0000https://tolgadimli.github.io/papers/paper3/We analyze the relationship between model performance and training dataset size for end-to-end autonomous driving.https://tolgadimli.github.io/papers/paper4/Fri, 15 Mar 2024 00:00:00 +0000https://tolgadimli.github.io/papers/paper4/We introduce gradient-norm–weighted averaging methods for distributed deep learning that pull workers toward flatter regions of the loss landscape. The proposed model-level and layer-level variants converge in both convex and non-convex settings and empirically achieve faster training, better optima, and reduced communication compared to prior work.https://tolgadimli.github.io/papers/paper5/Mon, 28 Aug 2023 00:00:00 +0000https://tolgadimli.github.io/papers/paper5/We propose an automated, bi-level document classification pipeline for regulatory text that first filters irrelevant documents using an ensemble of classical classifiers and then routes relevant documents to the appropriate departments using a transformer-based DocBERT model, significantly improving efficiency and accuracy over single-model approaches when evaluated on a large real-world corpus from Con Edison.https://tolgadimli.github.io/indcollab/data2/Tue, 15 Aug 2023 00:00:00 +0000https://tolgadimli.github.io/indcollab/data2/A bi-level document classification pipeline combining classical ML ensembles and transformer-based models to automate regulatory compliance management in large utility companies.https://tolgadimli.github.io/papers/paper6/Sun, 28 May 2023 00:00:00 +0000https://tolgadimli.github.io/papers/paper6/We introduce ConFER, a continual-learning benchmark for facial expression recognition that evaluates how well popular CL methods adapt to real-world, incremental data without catastrophic forgetting, demonstrating state-of-the-art performance across multiple FER datasets and highlighting the benefits of CL for robust affective computing.https://tolgadimli.github.io/location/Wed, 01 Sep 2021 00:00:00 +0000https://tolgadimli.github.io/location/Tolga Dimlioglu's office at NYU Tandon School of Engineering