Author name cluster

Georgios Kollias

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

6 papers

2 author rows

ICLR Conference 2025 Conference Paper

Large Language Models can Become Strong Self-Detoxifiers

Ching-Yun Ko
Pin-Yu Chen
Payel Das
Youssef Mroueh
Soham Dan
Georgios Kollias
Subhajit Chaudhury
Tejaswini Pedapati

Reducing the likelihood of generating harmful and toxic output is an essential task when aligning large language models (LLMs). Existing methods mainly rely on training an external reward model (i.e., another language model) or fine-tuning the LLM using self-generated data to influence the outcome. In this paper, we show that LLMs have the capability of self-detoxification without external reward model learning or retraining of the LM. We propose \textit{Self-disciplined Autoregressive Sampling (SASA)}, a lightweight controlled decoding algorithm for toxicity reduction of LLMs. SASA leverages the contextual representations from an LLM to learn linear subspaces from labeled data characterizing toxic v.s. non-toxic output in analytical forms. When auto-completing a response token-by-token, SASA dynamically tracks the margin of the current output to steer the generation away from the toxic subspace, by adjusting the autoregressive sampling strategy. Evaluated on LLMs of different scale and nature, namely Llama-3.1-Instruct (8B), Llama-2 (7B), and GPT2-L models with the RealToxicityPrompts, BOLD, and AttaQ benchmarks, SASA markedly enhances the quality of the generated sentences relative to the original models and attains comparable performance to state-of-the-art detoxification techniques, significantly reducing the toxicity level by only using the LLM's internal representations.

Details

TMLR Journal 2024 Journal Article

Directed Graph Transformers

Qitong Wang
Georgios Kollias
Vasileios Kalantzis
Naoki Abe
Mohammed J Zaki

In this paper, we address the problem of capturing graph directionality using transformers. Most existing graph transformers typically capture distances between graph nodes and do not take edge direction into account. This is a limiting assumption since many graph applications need to exploit sophisticated relationships in graph data, such as time, causality, or generic dependency constraints. We introduce a novel graph transformer architecture that explicitly takes into account the directionality between connected graph nodes. To achieve this, we make use of dual encodings to represent both potential roles, i.e., source or target, of each pair of vertices linked by a directed edge. These encodings are learned by leveraging the latent adjacency information extracted from a directional attention module, localized with $k$-hop neighborhood information. Extensive experiments on synthetic and real graph datasets show that our approach can have significant accuracy gains over previous graph transformer (GT) and graph neural network (GNN) approaches, providing state-of-the-art (SOTA) results on inherently directed graphs.

PDF Details

ICML Conference 2024 Conference Paper

Larimar: Large Language Models with Episodic Memory Control

Payel Das
Subhajit Chaudhury
Elliot Nelson
Igor Melnyk
Sarathkrishna Swaminathan
Sihui Dai
Aurélie C. Lozano
Georgios Kollias

Efficient and accurate updating of knowledge stored in Large Language Models (LLMs) is one of the most pressing research challenges today. This paper presents Larimar - a novel, brain-inspired architecture for enhancing LLMs with a distributed episodic memory. Larimar’s memory allows for dynamic, one-shot updates of knowledge without the need for computationally expensive re-training or fine-tuning. Experimental results on multiple fact editing benchmarks demonstrate that Larimar attains accuracy comparable to most competitive baselines, even in the challenging sequential editing setup, but also excels in speed—yielding speed-ups of 8-10x depending on the base LLM —as well as flexibility due to the proposed architecture being simple, LLM-agnostic, and hence general. We further provide mechanisms for selective fact forgetting, information leakage prevention, and input context length generalization with Larimar and show their effectiveness. Our code is available at https: //github. com/IBM/larimar.

Details

AAAI Conference 2022 Conference Paper

Directed Graph Auto-Encoders

Georgios Kollias
Vasileios Kalantzis
Tsuyoshi Ide
Aurélie Lozano
Naoki Abe

We introduce a new class of auto-encoders for directed graphs, motivated by a direct extension of the Weisfeiler-Leman algorithm to pairs of node labels. The proposed model learns pairs of interpretable latent representations for the nodes of directed graphs, and uses parameterized graph convolutional network (GCN) layers for its encoder and an asymmetric inner product decoder. Parameters in the encoder control the weighting of representations exchanged between neighboring nodes. We demonstrate the ability of the proposed model to learn meaningful latent embeddings and achieve superior performance on the directed link prediction task on several popular network datasets.

PDF Details

NeurIPS Conference 2021 Conference Paper

Cardinality-Regularized Hawkes-Granger Model

Tsuyoshi Ide
Georgios Kollias
Dzung Phan
Naoki Abe

We propose a new sparse Granger-causal learning framework for temporal event data. We focus on a specific class of point processes called the Hawkes process. We begin by pointing out that most of the existing sparse causal learning algorithms for the Hawkes process suffer from a singularity in maximum likelihood estimation. As a result, their sparse solutions can appear only as numerical artifacts. In this paper, we propose a mathematically well-defined sparse causal learning framework based on a cardinality-regularized Hawkes process, which remedies the pathological issues of existing approaches. We leverage the proposed algorithm for the task of instance-wise causal event analysis, where sparsity plays a critical role. We validate the proposed framework with two real use-cases, one from the power grid and the other from the cloud data center management domain.

PDF Details

ICML Conference 2021 Conference Paper

Projection techniques to update the truncated SVD of evolving matrices with applications

Vasileios Kalantzis
Georgios Kollias
Shashanka Ubaru
Athanasios N. Nikolakopoulos
Lior Horesh
Kenneth L. Clarkson

This submission considers the problem of updating the rank-$k$ truncated Singular Value Decomposition (SVD) of matrices subject to the addition of new rows and/or columns over time. Such matrix problems represent an important computational kernel in applications such as Latent Semantic Indexing and Recommender Systems. Nonetheless, the proposed framework is purely algebraic and targets general updating problems. The algorithm presented in this paper undertakes a projection viewpoint and focuses on building a pair of subspaces which approximate the linear span of the sought singular vectors of the updated matrix. We discuss and analyze two different choices to form the projection subspaces. Results on matrices from real applications suggest that the proposed algorithm can lead to higher accuracy, especially for the singular triplets associated with the largest modulus singular values. Several practical details and key differences with other approaches are also discussed.

Details