In self-supervised contrastive learning, negative pairs are typically constructed using an anchor image and a sample drawn from the entire dataset, excluding the anchor. However, this approach can result in the creation of negative pairs with similar semantics, referred to as "false negatives", leading to their embeddings being falsely pushed apart. To address this issue, we introduce GloFND, an optimization-based approach that automatically learns on the fly the threshold for each anchor data to identify its false negatives during training. In contrast to previous methods for false negative discovery, our approach globally detects false negatives across the entire dataset rather than locally within the mini-batch. Moreover, its per-iteration computation cost remains independent of the dataset size. Experimental results on image and image-text data demonstrate the effectiveness of the proposed method. Our implementation is available at https://github.com/vibalcam/GloFND .

Supervised fine-tuning (SFT) followed by preference optimization (PO) denoted by SFT$\rightarrow$PO has become the standard for improving pretrained large language models (LLMs), with PO demonstrating significant performance gains. However, PO methods rely on either human-labeled preference data or a strong reward model to generate preference data. Can we fine-tune LLMs without preference data or reward models while achieving competitive performance to SFT$\rightarrow$PO? We address this question by introducing Discriminative Fine-Tuning (DFT), a novel approach that eliminates the need for preference data. Unlike SFT, which employs a generative approach and overlooks negative data, DFT adopts a discriminative paradigm that that increases the probability of positive answers while suppressing potentially negative ones, shifting from token prediction to data prediction. Our contributions include: (i) a discriminative probabilistic framework for fine-tuning LLMs by explicitly modeling the discriminative likelihood of an answer among all possible outputs given an input; (ii) efficient algorithms to optimize this discriminative likelihood; and (iii) extensive experiments demonstrating DFT's effectiveness, achieving performance better than SFT and comparable to if not better than SFT$\rightarrow$PO. The code can be found at https://github.com/PenGuln/DFT.

Graph clustering is a fundamental task in graph mining where the goal is to partition nodes of a graph into disjoint clusters that have dense internal connections but are only sparsely connected to the rest of the graph. This has a wide variety of applications which include detecting communities in social networks [Fortunato, 2010], identifying related genes in biological networks based on gene expression profiles [Ben-Dor et al., 1999], and finding groups of pixels in an image that belong to the same object [Shi and Malik, 2000]. An idealized notion of a cluster in a graph is a set of nodes that is completely connected internally (i.e., a clique) while being completely disconnected from the rest of the graph. Cluster graph modification problems [Shamir et al., 2004] are a class of graph clustering objectives that seek to edit the edges in a graph as little as possible in order to achieve this idealized structure. One widely studied problem is correlation clustering [Bansal et al., 2004], which can be cast as adding or deleting a minimum number of edges to convert a graph into a disjoint union of cliques. This problem is also known as cluster editing.