Code Large Language Models (Code LLMs), such as Code llama and DeepSeek-Coder, have demonstrated exceptional performance in the code generation tasks. However, most existing models focus on the abilities of generating correct code, but often struggle with bug repair. We introduce a suit of methods to enhance LLM's SQL bug-fixing abilities. The methods are mainly consisted of two parts: A Progressive Dataset Construction (PDC) from scratch and Dynamic Mask Supervised Fine-tuning (DM-SFT). PDC proposes two data expansion methods from the perspectives of breadth first and depth first respectively. DM-SFT introduces an efficient bug-fixing supervised learning approach, which effectively reduce the total training steps and mitigate the "disorientation" in SQL code bug-fixing training. In our evaluation, the code LLM models trained with two methods have exceeds all current best performing model which size is much larger.

This paper introduces TIPS: Threat Actor Informed Prioritization using SecEncoder, a specialized language model for security. TIPS combines the strengths of both encoder and decoder language models to detect and prioritize compromised applications. By integrating threat actor intelligence, TIPS enhances the accuracy and relevance of its detections. Extensive experiments with a real-world benchmark dataset of applications demonstrate TIPS's high efficacy, achieving an F-1 score of 0.90 in identifying malicious applications. Additionally, in real-world scenarios, TIPS significantly reduces the backlog of investigations for security analysts by 87%, thereby streamlining the threat response process and improving overall security posture.

Impersonation tactics, such as app squatting and app cloning, have posed longstanding challenges in mobile app stores, where malicious actors exploit the names and reputations of popular apps to deceive users. With the rapid growth of Large Language Model (LLM) stores like GPT Store and FlowGPT, these issues have similarly surfaced, threatening the integrity of the LLM app ecosystem. In this study, we present the first large-scale analysis of LLM app squatting and cloning using our custom-built tool, LLMappCrazy. LLMappCrazy covers 14 squatting generation techniques and integrates Levenshtein distance and BERT-based semantic analysis to detect cloning by analyzing app functional similarities. Using this tool, we generated variations of the top 1000 app names and found over 5,000 squatting apps in the dataset. Additionally, we observed 3,509 squatting apps and 9,575 cloning cases across six major platforms. After sampling, we find that 18.7% of the squatting apps and 4.9% of the cloning apps exhibited malicious behavior, including phishing, malware distribution, fake content dissemination, and aggressive ad injection.

To enhance large language models (LLMs) for chemistry problem solving, several LLM-based agents augmented with tools have been proposed, such as ChemCrow and Coscientist. However, their evaluations are narrow in scope, leaving a large gap in understanding the benefits of tools across diverse chemistry tasks. To bridge this gap, we develop ChemAgent, an enhanced chemistry agent over ChemCrow, and conduct a comprehensive evaluation of its performance on both specialized chemistry tasks and general chemistry questions. Surprisingly, ChemAgent does not consistently outperform its base LLMs without tools. Our error analysis with a chemistry expert suggests that: For specialized chemistry tasks, such as synthesis prediction, we should augment agents with specialized tools; however, for general chemistry questions like those in exams, agents' ability to reason correctly with chemistry knowledge matters more, and tool augmentation does not always help.

Recent works have shown a surprising result: a small fraction of Large Language Model (LLM) parameter outliers are disproportionately important to the quality of the model. LLMs contain billions of parameters, so these small fractions, such as 0.01%, translate to hundreds of thousands of parameters. In this work, we present an even more surprising finding: Pruning as few as a single parameter can destroy an LLM's ability to generate text - increasing perplexity by 3 orders of magnitude and reducing zero-shot accuracy to guessing. We propose a data-free method for identifying such parameters, termed super weights, using a single forward pass through the model. We additionally find that these super weights induce correspondingly rare and large activation outliers, termed super activations. When preserved with high precision, super activations can improve simple round-to-nearest quantization to become competitive with state-of-the-art methods. For weight quantization, we similarly find that by preserving the super weight and clipping other weight outliers, round-to-nearest quantization can scale to much larger block sizes than previously considered. Large Language Models (LLMs) have been growing in size and capability at an unprecedented rate, enabling them to capture increasingly complex linguistic patterns across a wide range of tasks. However, with this increase in model scale, new and unexpected behaviors have emerged. Dettmers et al. (2022) discovered that once LLMs reach a certain scale, a small set of hidden state features contains outliers of exceptionally large magnitude.

Retrieval-augmented generation (RAG) is a promising method for addressing some of the memory-related challenges associated with Large Language Models (LLMs). Two separate systems form the RAG pipeline, the retriever and the reader, and the impact of each on downstream task performance is not well-understood. Here, we work towards the goal of understanding how retrievers can be optimized for RAG pipelines for common tasks such as Question Answering (QA). We conduct experiments focused on the relationship between retrieval and RAG performance on QA and attributed QA and unveil a number of insights useful to practitioners developing high-performance RAG pipelines. For example, lowering search accuracy has minor implications for RAG performance while potentially increasing retrieval speed and memory efficiency.

Monitoring public sentiment via social media is potentially helpful during health crises such as the COVID-19 pandemic. However, traditional frequency-based, data-driven neural network-based approaches can miss newly relevant content due to the evolving nature of language in a dynamically evolving environment. Human-curated symbolic knowledge sources, such as lexicons for standard language and slang terms, can potentially elevate social media signals in evolving language. We introduce a neurosymbolic method that integrates neural networks with symbolic knowledge sources, enhancing the detection and interpretation of mental health-related tweets relevant to COVID-19. Our method was evaluated using a corpus of large datasets (approximately 12 billion tweets, 2.5 million subreddit data, and 700k news articles) and multiple knowledge graphs. This method dynamically adapts to evolving language, outperforming purely data-driven models with an F1 score exceeding 92\%. This approach also showed faster adaptation to new data and lower computational demands than fine-tuning pre-trained large language models (LLMs). This study demonstrates the benefit of neurosymbolic methods in interpreting text in a dynamic environment for tasks such as health surveillance.

As large language models (LLMs) grow more powerful, ensuring their safety against misuse becomes crucial. While researchers have focused on developing robust defenses, no method has yet achieved complete invulnerability to attacks. We propose an alternative approach: instead of seeking perfect adversarial robustness, we develop rapid response techniques to look to block whole classes of jailbreaks after observing only a handful of attacks. To study this setting, we develop RapidResponseBench, a benchmark that measures a defense's robustness against various jailbreak strategies after adapting to a few observed examples. We evaluate five rapid response methods, all of which use jailbreak proliferation, where we automatically generate additional jailbreaks similar to the examples observed. Our strongest method, which fine-tunes an input classifier to block proliferated jailbreaks, reduces attack success rate by a factor greater than 240 on an in-distribution set of jailbreaks and a factor greater than 15 on an out-of-distribution set, having observed just one example of each jailbreaking strategy. Moreover, further studies suggest that the quality of proliferation model and number of proliferated examples play an key role in the effectiveness of this defense. Overall, our results highlight the potential of responding rapidly to novel jailbreaks to limit LLM misuse. As Large Language Models (LLMs) become more capable, they pose greater misuse risks. Indeed, the potential for catastrophic misuse of LLMs has motivated AI labs to make public commitments to developing safeguards to minimize the risk of such misuse (Anthropic, 2023; OpenAI, 2023). Additionally, such concerns have motivated substantial effort from the research community to defend against jailbreaks, which are techniques that extract harmful information from LLMs trained to be helpful, harmless, and honest (Bai et al., 2022b; Xie et al., 2023; Xu et al., 2024). Despite ongoing research, ensuring that large language models (LLMs) are robustly resistant to jailbreaking remains an unsolved challenge (Hendrycks et al., 2021; Ziegler et al., 2022). Even state-of-the-art methods that substantially improve robustness, such as representation rerouting (Zou et al., 2024), have been publicly broken within hours of release. The situation could worryingly parallel that of adversarial robustness in computer vision, where new defenses are often defeated by attacks available before their development with proper tuning (Tramer et al., 2020). Indeed, in computer vision, a decade of work and thousands of papers have yielded "limited progress" (Carlini, 2024).

We introduce two complementary techniques for efficient adaptive optimization that reduce memory requirements while accelerating training of large-scale neural networks. The first technique, Subset-Norm adaptive step size, generalizes AdaGrad-Norm and AdaGrad(-Coordinate) by reducing the second moment term's memory footprint from $O(d)$ to $O(\sqrt{d})$ through step-size sharing, where $d$ is the model size. For non-convex smooth objectives under coordinate-wise sub-gaussian gradient noise, we prove a noise-adapted high-probability convergence guarantee showing improved dimensional dependence over existing methods. Our second technique, Subspace-Momentum, reduces the momentum state's memory footprint by operating in a low-dimensional subspace while applying standard SGD in the orthogonal complement. We establish high-probability convergence rates under similar relaxed assumptions. Empirical evaluation on LLaMA models from 60M to 1B parameters demonstrates the effectiveness of our methods, where combining subset-norm with subspace-momentum achieves Adam's validation perplexity in approximately half the training tokens (6.8B vs 13.1B) while using only 20% of the Adam's optimizer-states memory footprint and requiring minimal additional hyperparameter tuning.

To predict upcoming text, language models must in some cases retrieve in-context information verbatim. In this report, we investigated how the ability of language models to retrieve arbitrary in-context nouns developed during training (across time) and as language models trained on the same dataset increase in size (across scale). We then asked whether learning of in-context retrieval correlates with learning of more challenging zero-shot benchmarks. Furthermore, inspired by semantic effects in human short-term memory, we evaluated the retrieval with respect to a major semantic component of target nouns, namely whether they denote a concrete or abstract entity, as rated by humans. We show that verbatim in-context retrieval developed in a sudden transition early in the training process, after about 1% of the training tokens. This was observed across model sizes (from 14M and up to 12B parameters), and the transition occurred slightly later for the two smallest models. We further found that the development of verbatim in-context retrieval is positively correlated with the learning of zero-shot benchmarks. Around the transition point, all models showed the advantage of retrieving concrete nouns as opposed to abstract nouns. In all but two smallest models, the advantage dissipated away toward the end of training.