Meta-training, which fine-tunes the language model (LM) on various downstream tasks by maximizing the likelihood of the target label given the task instruction and input instance, has improved the zero-shot task generalization performance. However, meta-trained LMs still struggle to generalize to challenging tasks containing novel labels unseen during meta-training. In this paper, we propose Flipped Learning, an alternative method of meta-training which trains the LM to generate the task instruction given the input instance and label. During inference, the LM trained with Flipped Learning, referred to as Flipped, selects the label option that is most likely to generate the task instruction. On 14 tasks of the BIG-bench benchmark, the 11B-sized Flipped outperforms zero-shot T0-11B and even a 16 times larger 3-shot GPT-3 (175B) on average by 8.4% and 9.7% points, respectively. Flipped gives particularly large improvements on tasks with unseen labels, outperforming T0-11B by up to +20% average F1 score. This indicates that the strong task generalization of Flipped comes from improved generalization to novel labels. We release our code at https://github.com/seonghyeonye/Flipped-Learning.

Pretrained Language Models (LMs) memorize a vast amount of knowledge during initial pretraining, including information that may violate the privacy of personal lives and identities. Previous work addressing privacy issues for language models has mostly focused on data preprocessing and differential privacy methods, both requiring re-training the underlying LM. We propose knowledge unlearning as an alternative method to reduce privacy risks for LMs post hoc. We show that simply performing gradient ascent on target token sequences is effective at forgetting them with little to no degradation of general language modeling performances for larger LMs; it sometimes even substantially improves the underlying LM with just a few iterations. We also find that sequential unlearning is better than trying to unlearn all the data at once and that unlearning is highly dependent on which kind of data (domain) is forgotten. By showing comparisons with a previous data preprocessing method and a decoding method known to mitigate privacy risks for LMs, we show that unlearning can give a stronger empirical privacy guarantee in scenarios where the data vulnerable to extraction attacks are known a priori while being much more efficient and robust. We release the code and dataset needed to replicate our results at https://github.com/joeljang/knowledge-unlearning.

先前的工作表明，语言模型（LMS）的大小（LMS）与它们在不同下游NLP任务上的零拍摄性能之间存在缩放定律。在这项工作中，我们表明，在用否定提示的任务评估大型LM时，这种现象并不存在，而是显示了逆缩放定律。我们对（1）验证的LMS（OPT＆GPT -3）的否定提示评估了9个不同的任务，该任务的不同尺寸（125m -175b），（2）LMS进一步预处理以推广到新颖的提示（指令），（3）提供的LMS，（3）LMS。示例很少，（4）LMS专门针对否定的提示进行了微调；所有LM类型在否定的提示上的表现较差，并在比较原始提示和否定提示的平均得分时显示人类绩效之间的巨大性能差距。通过强调现有LMS和方法的关键局限，我们敦促社区开发开发实际遵循给定指示的LMS的新方法。我们提供代码和数据集，以探索https://github.com/joeljang/negated-prompts-for-llms的否定提示。

最近的作品表明，将提示附加到输入方面有效，可以在调节语言模型（LM）上执行特定任务。但是，提示在推理过程中始终包含在输入文本中，从而产生了实质性的计算和内存开销。同样，目前尚无直接使用提示的方法，其提示比LMS的最大输入长度更长，而不会在推理过程中产生额外费用。我们提出了提示注射（PI），这是一种将提示注入LM参数的新型公式，是将固定提示附加到输入的有效替代方案。我们表明，在有固定提示的情况下，PI的总拖鞋的效率可能比以前的方法高280倍。我们进一步探讨了PI的方法论，并通过角色依赖的对话，语义解析和零射击学习显示了有希望的结果。通过这些探索，我们表明PI可以成为调节语言模型的有前途的方向，尤其是在具有长长和固定提示的情况下。

Dense prediction tasks such as segmentation and detection of pathological entities hold crucial clinical value in the digital pathology workflow. However, obtaining dense annotations on large cohorts is usually tedious and expensive. Contrastive learning (CL) is thus often employed to leverage large volumes of unlabeled data to pre-train the backbone network. To boost CL for dense prediction, some studies have proposed variations of dense matching objectives in pre-training. However, our analysis shows that employing existing dense matching strategies on histopathology images enforces invariance among incorrect pairs of dense features and, thus, is imprecise. To address this, we propose a precise location-based matching mechanism that utilizes the overlapping information between geometric transformations to precisely match regions in two augmentations. Extensive experiments on two pretraining datasets (TCGA-BRCA, NCT-CRC-HE) and three downstream datasets (GlaS, CRAG, BCSS) highlight the superiority of our method in semantic and instance segmentation tasks. Our method outperforms previous dense matching methods by up to 7.2 % in average precision for detection and 5.6 % in average precision for instance segmentation tasks. Additionally, by using our matching mechanism in the three popular contrastive learning frameworks, MoCo-v2, VICRegL and ConCL, the average precision in detection is improved by 0.7 % to 5.2 % and the average precision in segmentation is improved by 0.7 % to 4.0 %, demonstrating its generalizability.

Inspired by the recent success of Transformers for Natural Language Processing and vision Transformer for Computer Vision, many researchers in the medical imaging community have flocked to Transformer-based networks for various main stream medical tasks such as classification, segmentation, and estimation. In this study, we analyze, two recently published Transformer-based network architectures for the task of multimodal head-and-tumor segmentation and compare their performance to the de facto standard 3D segmentation network - the nnU-Net. Our results showed that modeling long-range dependencies may be helpful in cases where large structures are present and/or large field of view is needed. However, for small structures such as head-and-neck tumor, the convolution-based U-Net architecture seemed to perform well, especially when training dataset is small and computational resource is limited.

The proliferation of automatic faithfulness metrics for summarization has produced a need for benchmarks to evaluate them. While existing benchmarks measure the correlation with human judgements of faithfulness on model-generated summaries, they are insufficient for diagnosing whether metrics are: 1) consistent, i.e., decrease as errors are introduced into a summary, 2) effective on human-written texts, and 3) sensitive to different error types (as summaries can contain multiple errors). To address these needs, we present a benchmark of unfaithful minimal pairs (BUMP), a dataset of 889 human-written, minimally different summary pairs, where a single error (from an ontology of 7 types) is introduced to a summary from the CNN/DailyMail dataset to produce an unfaithful summary. We find BUMP complements existing benchmarks in a number of ways: 1) the summaries in BUMP are harder to discriminate and less probable under SOTA summarization models, 2) BUMP enables measuring the consistency of metrics, and reveals that the most discriminative metrics tend not to be the most consistent, 3) BUMP enables the measurement of metrics' performance on individual error types and highlights areas of weakness for future work.

In recent years, generative models have undergone significant advancement due to the success of diffusion models. The success of these models is often attributed to their use of guidance techniques, such as classifier and classifier-free methods, which provides effective mechanisms to trade-off between fidelity and diversity. However, these methods are not capable of guiding a generated image to be aware of its geometric configuration, e.g., depth, which hinders the application of diffusion models to areas that require a certain level of depth awareness. To address this limitation, we propose a novel guidance approach for diffusion models that uses estimated depth information derived from the rich intermediate representations of diffusion models. To do this, we first present a label-efficient depth estimation framework using the internal representations of diffusion models. At the sampling phase, we utilize two guidance techniques to self-condition the generated image using the estimated depth map, the first of which uses pseudo-labeling, and the subsequent one uses a depth-domain diffusion prior. Experiments and extensive ablation studies demonstrate the effectiveness of our method in guiding the diffusion models toward geometrically plausible image generation. Project page is available at https://ku-cvlab.github.io/DAG/.

How can we accurately identify new memory workloads while classifying known memory workloads? Verifying DRAM (Dynamic Random Access Memory) using various workloads is an important task to guarantee the quality of DRAM. A crucial component in the process is open-set recognition which aims to detect new workloads not seen in the training phase. Despite its importance, however, existing open-set recognition methods are unsatisfactory in terms of accuracy since they fail to exploit the characteristics of workload sequences. In this paper, we propose Acorn, an accurate open-set recognition method capturing the characteristics of workload sequences. Acorn extracts two types of feature vectors to capture sequential patterns and spatial locality patterns in memory access. Acorn then uses the feature vectors to accurately classify a subsequence into one of the known classes or identify it as the unknown class. Experiments show that Acorn achieves state-of-the-art accuracy, giving up to 37% points higher unknown class detection accuracy while achieving comparable known class classification accuracy than existing methods.

The number of international benchmarking competitions is steadily increasing in various fields of machine learning (ML) research and practice. So far, however, little is known about the common practice as well as bottlenecks faced by the community in tackling the research questions posed. To shed light on the status quo of algorithm development in the specific field of biomedical imaging analysis, we designed an international survey that was issued to all participants of challenges conducted in conjunction with the IEEE ISBI 2021 and MICCAI 2021 conferences (80 competitions in total). The survey covered participants' expertise and working environments, their chosen strategies, as well as algorithm characteristics. A median of 72% challenge participants took part in the survey. According to our results, knowledge exchange was the primary incentive (70%) for participation, while the reception of prize money played only a minor role (16%). While a median of 80 working hours was spent on method development, a large portion of participants stated that they did not have enough time for method development (32%). 25% perceived the infrastructure to be a bottleneck. Overall, 94% of all solutions were deep learning-based. Of these, 84% were based on standard architectures. 43% of the respondents reported that the data samples (e.g., images) were too large to be processed at once. This was most commonly addressed by patch-based training (69%), downsampling (37%), and solving 3D analysis tasks as a series of 2D tasks. K-fold cross-validation on the training set was performed by only 37% of the participants and only 50% of the participants performed ensembling based on multiple identical models (61%) or heterogeneous models (39%). 48% of the respondents applied postprocessing steps.