Action recognition models demonstrate strong generalization, but can they effectively transfer high-level motion concepts across diverse contexts, even within similar distributions? For example, can a model recognize the broad action "punching" when presented with an unseen variation such as "punching person"? To explore this, we introduce a motion transferability framework with three datasets: (1) Syn-TA, a synthetic dataset with 3D object motions; (2) Kinetics400-TA; and (3) Something-Something-v2-TA, both adapted from natural video datasets. We evaluate 13 state-of-the-art models on these benchmarks and observe a significant drop in performance when recognizing high-level actions in novel contexts. Our analysis reveals: 1) Multimodal models struggle more with fine-grained unknown actions than with coarse ones; 2) The bias-free Syn-TA proves as challenging as real-world datasets, with models showing greater performance drops in controlled settings; 3) Larger models improve transferability when spatial cues dominate but struggle with intensive temporal reasoning, while reliance on object and background cues hinders generalization. We further explore how disentangling coarse and fine motions can improve recognition in temporally challenging datasets. We believe this study establishes a crucial benchmark for assessing motion transferability in action recognition. Datasets and relevant code: https://github.com/raiyaan-abdullah/Motion-Transfer.

Existing long video retrieval systems are trained and tested in the paragraph-to- video retrieval regime, where ev-ery long video is described by a single long paragraph. This neglects the richness and variety of possible valid de-scriptions of a video, which could range anywhere from moment-by-moment detail to a single phrase summary. To provide a more thorough evaluation of the capabilities of long video retrieval systems, we propose a pipeline that leverages state-of-the-art large language models to care-fully generate a diverse set of synthetic captions for long videos. We validate this pipeline's fidelity via rigorous hu-man inspection. We use synthetic captionsfrom this pipeline to perform a benchmark of a representative set of video language models using long video datasets, and show that the models struggle on shorter captions. We show that finetuning on this data can both mitigate these issues (+2.8% R@ 1 over SOTA on ActivityNet with diverse captions), and even improve performance on standard paragraph-to-video re-trieval (+ 1.0% R@1 on ActivityNet). We also use synthetic data from our pipeline as query expansion in the zero-shot setting (+3.4% R@ 1 on ActivityNet). We derive insights by analyzing failure cases for retrieval with short captions.

We present DRESS , a large vision language model (LVLM) that innovatively exploits Natural Language feedback (NLF) from Large Language Models to enhance its alignment and interactions by addressing two key limitations in the state-of-the-art LVLMs. First, prior LVLMs generally rely only on the instruction finetuning stage to enhance alignment with human preferences. Without incorporating extra feedback, they are still prone to generate unhelpful, hallucinated, or harmful responses. Second, while the visual instruction tuning data is generally structured in a multi-turn dialogue format, the connections and dependencies among consecutive conversational turns are weak. This reduces the capacity for effective multi-turn interactions. To tackle these, we propose a novel categorization of the NLF into two key types: critique and refinement. The critique NLF identifies the strengths and weaknesses of the responses and is used to align the LVLMs with human preferences. The refinement NLF offers concrete suggestions for improvement and is adopted to improve the interaction ability of the LVLMs- which focuses on LVLMs' ability to refine responses by incorporating feedback in multi-turn interactions. To address the non-differentiable nature of NLF, we generalize conditional reinforcement learning for training. Our experimental results demonstrate that DRESS can generate more helpful (9.76%), honest (11.52%), and harmless (21.03%) responses, and more effectively learn from feedback during multi-turn interactions compared to SOTA LVLMs.

Yangyi Chen, Karan Sikka, Michael Cogswell, Heng Ji, Ajay Divakaran. Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers). 2024.

Can we develop visually grounded dialog agents that can efficiently adapt to new tasks without forgetting how to talk to people? Such agents could leverage a larger variety of existing data to generalize to a new task, minimizing expensive data collection and annotation. In this work, we study a setting we call "Dialog without Dialog", which requires agents to develop visually grounded dialog models that can adapt to new tasks without language level supervision. By factorizing intention and language, our model minimizes linguistic drift after fine-tuning for new tasks. We present qualitative results, automated metrics, and human studies that all show our model can adapt to new tasks and maintain language quality. Baselines either fail to perform well at new tasks or experience language drift, becoming unintelligible to humans. Code has been made available at: https://github.com/mcogswell/dialogwithoutdialog.

A single image captures the appearance and position of multiple entities in a scene as well as their complex interactions. As a consequence, natural language grounded in visual contexts tends to be diverse---with utterances differing as focus shifts to specific objects, interactions, or levels of detail. Recently, neural sequence models such as RNNs and LSTMs have been employed to produce visually-grounded language. Beam Search, the standard work-horse for decoding sequences from these models, is an approximate inference algorithm that decodes the top-B sequences in a greedy left-to-right fashion. In practice, the resulting sequences are often minor rewordings of a common utterance, failing to capture the multimodal nature of source images. To address this shortcoming, we propose Diverse Beam Search (DBS), a diversity promoting alternative to BS for approximate inference. DBS produces sequences that are significantly different from each other by incorporating diversity constraints within groups of candidate sequences during decoding; moreover, it achieves this with minimal computational or memory overhead. We demonstrate that our method improves both diversity and quality of decoded sequences over existing techniques on two visually-grounded language generation tasks---image captioning and visual question generation---particularly on complex scenes containing diverse visual content. We also show similar improvements at language-only machine translation tasks, highlighting the generality of our approach.

We propose a technique for producing `visual explanations' for decisions from a large class of Convolutional Neural Network (CNN)-based models, making them more transparent. Our approach - Gradient-weighted Class Activation Mapping (Grad-CAM), uses the gradients of any target concept (say logits for `dog' or even a caption), flowing into the final convolutional layer to produce a coarse localization map highlighting the important regions in the image for predicting the concept. Unlike previous approaches, Grad- CAM is applicable to a wide variety of CNN model-families: (1) CNNs with fully-connected layers (e.g. VGG), (2) CNNs used for structured outputs (e.g. captioning), (3) CNNs used in tasks with multi-modal inputs (e.g. visual question answering) or reinforcement learning, without architectural changes or re-training. We combine Grad-CAM with existing fine-grained visualizations to create a high-resolution class-discriminative visualization, Guided Grad-CAM, and apply it to image classification, image captioning, and visual question answering (VQA) models, including ResNet-based architectures. In the context of image classification models, our visualizations (a) lend insights into failure modes of these models (showing that seemingly unreasonable predictions have reasonable explanations), (b) outperform previous methods on the ILSVRC-15 weakly-supervised localization task, (c) are more faithful to the underlying model, and (d) help achieve model generalization by identifying dataset bias. For image captioning and VQA, our visualizations show even non-attention based models can localize inputs. Finally, we design and conduct human studies to measure if Grad-CAM explanations help users establish appropriate trust in predictions from deep networks and show that Grad-CAM helps untrained users successfully discern a `stronger' deep network from a `weaker' one even when both make identical predictions. Our code is available at https: //github.com/ramprs/grad-cam/ along with a demo on CloudCV [2] and video at youtu.be/COjUB9Izk6E.

Many practical perception systems exist within larger processes which often include interactions with users or additional components that are capable of evaluating the quality of predicted solutions. In these contexts, it is beneficial to provide these oracle mechanisms with multiple highly likely hypotheses rather than a single prediction. In this work, we pose the task of producing multiple outputs as a learning problem over an ensemble of deep networks -- introducing a novel stochastic gradient descent based approach to minimize the loss with respect to an oracle. Our method is simple to implement, agnostic to both architecture and loss function, and parameter-free. Our approach achieves lower oracle error compared to existing methods on a wide range of tasks and deep architectures. We also show qualitatively that solutions produced from our approach often provide interpretable representations of task ambiguity.

At SIGCSE 2002, Michael Goldwasser suggested a strategy for adding software testing practices to programming courses by requiring students to turn in tests along with their solutions, and then running every student's tests against every other student's program. This approach provides a much more robust environment for assessing the quality of student-written tests, and also provides more thorough testing of student solutions. Although software testing is included as a regular part of many more programming courses today, the all-pairs model of executing tests is still a rarity. This is because student-written tests, such as JUnit tests written for Java programs, are now more commonly written in the form of program code themselves, and they may depend on virtually any aspect of their author's own solution. These dependencies may keep one student's tests from even compiling against another student's program. This paper discusses the problem and presents a novel solution for Java that uses bytecode rewriting to transform a student's tests into a form that uses reflection to run against any other solution, regardless of any compile-time dependencies that may have been present in the original tests. Results of applying this technique to two assignments, encompassing 147 student programs and 240,158 individual test case runs, shows the feasibility of the approach and provides some insight into the quality of both student tests and student programs. An analysis of these results is presented.

The boundary scan test methodology is becoming an increasingly important approach for testing chips, modules and boards. Commercial boundary scan verification tools are now available which provide a system of checks not only for IEEE 1149.1 but other methodologies such as IBM Boundary Scan. A key factor in the effectiveness of boundary scan verification systems is found in the accuracy and flexibility of companion analysis tools used to correlate the violated boundary scan rule with the subject logic structure causing the violation. This paper presents the design and deployment of a graphical system for pinpointing sources of boundary scan rules violations. The paper begins with a cursory review of boundary scan methodologies including IEEE 1149.1 and IBM Boundary Scan. This is followed by a brief presentation of the boundary scan verification process used in IBM's TestBench tool. The body of the paper is focused on boundary scan verification rules and the associated message analysis. The paper concludes with future plans under consideration to improve both the reach and usability of graphical message analysis for boundary scan verification.

Signature Based Test Generation continues to grow in importance as VLSI circuits cross and leap beyond the multi-million gate mark. The Logic Built-In Self Test (LBIST) methodology is a signature analyst's based test generation strategy in wide spread use today. Since the LBIST hardware forms the basis for the generation of test patterns and failure-mode diagnostics, it is paramount that the LBIST test structures conform to LBIST methodology requirements. If the LBIST structures are modeled incorrectly, then pattern mismatches may occur at the tester. This paper describes a set of production level algorithms and procedures used to validate the LBIST structures used in the LBIST methodology. The validation includes the identification and verification of individual components of the LBIST structure. These algorithms are based on the test structure validation processes designed within IBM's TestBench test generation system.

Test Structure Verification (TSV) is the process used to assess the conformance of a circuit to a set of predefine Design For Testability (DFT) rules. Test generation algorithms are typically optimized for circuits that are highly testable, so Test Structure Verification (TSV) ultimately indicates the overall testability of a circuit. Logical BIST (Built In Self Test) is a methodology based on signature analysis which produces patterns used to identify the correctness of the manufactured circuit via specially designed BIST hardware. This paper describes a set of proven production level procedures used to identify and verify the test structure and behavior of BIST hardware. These algorithms are based on the TSV implementation of IBM's TestBench test generation system.

The first suite of Power PC processors such as the PowerPC 601 and PowerPC 603 are making their way into both desktop and special purpose hardware applications. Additional PowerPC processors are being developed including processors designed for the high-end server and workstation market as well as those which function efficiently in real-time embedded control applications. In this paper, the PowerPC visual simulator (PVS) is introduced. Following this, several PowerPC application scenarios are described. Each of these scenarios focuses on a common software analysis or testing problem. PowerPC solutions to these problems are explored using the PVS tool.>