Abuse reporting services collect reports about abuse victims have suffered. Accurate classification of the submitted reports is fundamental to analyzing the prevalence and financial impact of different abuse types (e.g., sextortion, investment, romance). Current classification approaches are problematic because they require the reporter to select the abuse type from a list, assuming the reporter has the necessary experience for the classification, which we show is frequently not the case, or require manual classification by analysts, which does not scale. To address these issues, this paper presents a novel approach to classify cryptocurrency abuse reports automatically. We first build a taxonomy of 19 frequently reported abuse types. Given as input the textual description written by the reporter, our classifier leverages a large language model (LLM) to interpret the text and assign it an abuse type in our taxonomy. We collect 290K cryptocurrency abuse reports from two popular reporting services: BitcoinAbuse and BBB's ScamTracker. We build ground truth datasets for 20K of those reports and use them to evaluate three designs for our LLM-based classifier and four LLMs, as well as a supervised ML classifier used as a baseline. Our LLM-based classifier achieves a precision of 0.92, a recall of 0.87, and an F1 score of 0.89, compared to an F1 score of 0.55 for the baseline. We demonstrate our classifier in two applications: providing financial loss statistics for fine-grained abuse types and generating tagged addresses for cryptocurrency analysis platforms.

Since the release of ChatGPT, large language models (LLMs) have demonstrated remarkable capabilities across various domains. A key challenge in developing these general capabilities is efficiently sourcing diverse, high-quality data. This becomes especially critical in reasoning-related tasks with sandbox checkers, such as math or code, where the goal is to generate correct solutions to specific problems with higher probability. In this work, we introduce Flaming-hot Initiation with Regular Execution (FIRE) sampling, a simple yet highly effective method to efficiently find good responses. Our empirical findings show that FIRE sampling enhances inference-time generation quality and also benefits training in the alignment stage. Furthermore, we explore how FIRE sampling improves performance by promoting diversity and analyze the impact of employing FIRE at different positions within a response.

Language models are now deployed in a wide variety of user-facing applications, often for specific purposes like answering questions about documentation or acting as coding assistants. As these models are intended for particular purposes, they should not be able to answer irrelevant queries like requests for poetry or questions about physics, or even worse, queries that can only be answered by humans like sensitive company policies. Instead we would like them to only answer queries corresponding to desired behavior and refuse all other requests, which we refer to as scoping. We find that, despite the use of system prompts, two representative language models can be poorly scoped and respond to queries they should not be addressing. We then conduct a comprehensive empirical evaluation of methods which could be used for scoping the behavior of language models. Among many other results, we show that a recently-proposed method for general alignment, Circuit Breakers (CB), can be adapted to scope language models to very specific tasks like sentiment analysis or summarization or even tasks with finer-grained scoping (e.g. When compared to standard methods like fine-tuning or preference learning, CB is more robust both for out of distribution tasks, and to adversarial prompting techniques. We also show that layering SFT and CB together often results in the best of both worlds: improved performance only on relevant queries, while rejecting irrelevant ones. In the past few years Large Language Models have exploded into the popular conscience. One major recent addition is the "alignment" process through Reinforcement Learning with Human Feedback (RLHF) (Christiano et al., 2017; Ouyang et al., 2022), which has made the current generation of language models much less likely to emit toxic content than previous generations (Wolf et al., 2017), and thus much more acceptable for general use. As a result, many businesses and individuals feel more comfortable using these technologies than they would be in the past. As a result, we have generally capable language models which refuse to answer toxic or dangerous queries, but it is still difficult to deploy these language models. Even though they may not emit toxic content as often, they still will happily answer any question, irrelevant or not. This becomes a problem when we wish to deploy language models as products in specific contexts: e.g. While language models have general language capability, there is still a need to scope them for specific uses. David Yunis is a PhD student at the Toyota Technological Institute at Chicago. Work was performed during an internship at IBM. Arrows indicate the direction of best performance.

Large language models (LLMs; Jiang et al., 2023; Touvron et al., 2023; Dubey et al., 2024; Achiam et al., 2023) have demonstrated remarkable performance on numerous downstream natural language processing (NLP) tasks, leading to their widespread integration into various decision-making processes (Bommasani et al., 2021; Band et al., 2024; Zhou et al., 2024). However, even with significant increases in model size and the expansion of training datasets, it remains infeasible for LLMs to encode all possible knowledge within their parameters. As a result, the outputs produced by LLMs may not consistently be reliable for important human decision-making processes, potentially overlooking key or hidden details. Additionally, LLMs frequently provide inaccurate or misleading information with a high degree of confidence, a phenomenon referred to as hallucination (Zhuo et al., 2023; Papamarkou et al., 2024), which can lead humans to make flawed decisions. In addition, Zhou et al. (2024) have empirically demonstrated that human users often over-rely on LLM outputs during decision-making processes, and this over-reliance tends to increase in proportion to the model's confidence.

Education is being transformed by rapid advances in Artificial Intelligence (AI), including emerging Generative Artificial Intelligence (GAI). Such technology can significantly support academics and students by automating monotonous tasks and making personalised suggestions. However, despite the potential of the technology, there are significant concerns regarding AI misuse, particularly by students in assessments. There are two schools of thought: one advocates for a complete ban on it, while the other views it as a valuable educational tool, provided it is governed by a robust usage policy. This contradiction clearly indicates a major policy gap in academic practices, and new policies are required to uphold academic standards while enabling staff and students to benefit from technological advancements. We surveyed 117 academics from three countries (UK, UAE, and Iraq), and identified that most academics retain positive opinions regarding AI in education. For example, the majority of experienced academics do not favour complete bans, and they see the potential benefits of AI for students, teaching staff, and academic institutions. Importantly, academics specifically identified the particular benefits of AI for autonomous assessment (71.79% of respondents agreed). Therefore, for the first time, we propose a novel AI framework for autonomously evaluating students' work (e.g., reports, coursework, etc.) and automatically assigning grades based on their knowledge and in-depth understanding of the submitted content. The survey results further highlight a significant lack of awareness of modern AI-based tools (e.g., ChatGPT) among experienced academics, a gap that must be addressed to uphold educational standards.

As the scale of training corpora for large language models (LLMs) grows, model developers become increasingly reluctant to disclose details on their data. This lack of transparency poses challenges to scientific evaluation and ethical deployment. Recently, pretraining data detection approaches, which infer whether a given text was part of an LLM's training data through black-box access, have been explored. The Min-K\% Prob method, which has achieved state-of-the-art results, assumes that a non-training example tends to contain a few outlier words with low token probabilities. However, the effectiveness may be limited as it tends to misclassify non-training texts that contain many common words with high probabilities predicted by LLMs. To address this issue, we introduce a divergence-based calibration method, inspired by the divergence-from-randomness concept, to calibrate token probabilities for pretraining data detection. We compute the cross-entropy (i.e., the divergence) between the token probability distribution and the token frequency distribution to derive a detection score. We have developed a Chinese-language benchmark, PatentMIA, to assess the performance of detection approaches for LLMs on Chinese text. Experimental results on English-language benchmarks and PatentMIA demonstrate that our proposed method significantly outperforms existing methods. Our code and PatentMIA benchmark are available at \url{https://github.com/zhang-wei-chao/DC-PDD}.

Do large language models (LLMs) solve reasoning tasks by learning robust generalizable algorithms, or do they memorize training data? To investigate this question, we use arithmetic reasoning as a representative task. Using causal analysis, we identify a subset of the model (a circuit) that explains most of the model's behavior for basic arithmetic logic and examine its functionality. By zooming in on the level of individual circuit neurons, we discover a sparse set of important neurons that implement simple heuristics. Each heuristic identifies a numerical input pattern and outputs corresponding answers. We hypothesize that the combination of these heuristic neurons is the mechanism used to produce correct arithmetic answers. To test this, we categorize each neuron into several heuristic types-such as neurons that activate when an operand falls within a certain range-and find that the unordered combination of these heuristic types is the mechanism that explains most of the model's accuracy on arithmetic prompts. Finally, we demonstrate that this mechanism appears as the main source of arithmetic accuracy early in training. Overall, our experimental results across several LLMs show that LLMs perform arithmetic using neither robust algorithms nor memorization; rather, they rely on a "bag of heuristics".

Large language models (LLMs) have demonstrated remarkable performance across various downstream tasks. However, the high computational and memory requirements of LLMs are a major bottleneck. To address this, parameter-efficient fine-tuning (PEFT) methods such as low-rank adaptation (LoRA) have been proposed to reduce computational costs while ensuring minimal loss in performance. Additionally, knowledge distillation (KD) has been a popular choice for obtaining compact student models from teacher models. In this work, we present KD-LoRA, a novel fine-tuning method that combines LoRA with KD. Our results demonstrate that KD-LoRA achieves performance comparable to full fine-tuning (FFT) and LoRA while significantly reducing resource requirements. Specifically, KD-LoRA retains 98% of LoRA's performance on the GLUE benchmark, while being 40% more compact. Additionally, KD-LoRA reduces GPU memory usage by 30% compared to LoRA, while decreasing inference time by 30% compared to both FFT and LoRA. We evaluate KD-LoRA across three encoder-only models: BERT, RoBERTa, and DeBERTaV3. Code is available at https://github.com/rambodazimi/KD-LoRA.

Large language models can produce highly fluent paraphrases while retaining much of the original meaning. While this capability has a variety of helpful applications, it may also be abused by bad actors, for example to plagiarize content or to conceal their identity. This motivates us to consider the problem of paraphrase inversion: given a paraphrased document, attempt to recover the original text. To explore the feasibility of this task, we fine-tune paraphrase inversion models, both with and without additional author-specific context to help guide the inversion process. We explore two approaches to author-specific inversion: one using in-context examples of the target author's writing, and another using learned style representations that capture distinctive features of the author's style. We show that, when starting from paraphrased machine-generated text, we can recover significant portions of the document using a learned inversion model. When starting from human-written text, the variety of source writing styles poses a greater challenge for invertability. However, even when the original tokens can't be recovered, we find the inverted text is stylistically similar to the original, which significantly improves the performance of plagiarism detectors and authorship identification systems that rely on stylistic markers.

This paper introduces NeuGPT, a groundbreaking multi-modal language generation model designed to harmonize the fragmented landscape of neural recording research. Traditionally, studies in the field have been compartmentalized by signal type, with EEG, MEG, ECoG, SEEG, fMRI, and fNIRS data being analyzed in isolation. Recognizing the untapped potential for cross-pollination and the adaptability of neural signals across varying experimental conditions, we set out to develop a unified model capable of interfacing with multiple modalities. Drawing inspiration from the success of pre-trained large models in NLP, computer vision, and speech processing, NeuGPT is architected to process a diverse array of neural recordings and interact with speech and text data. Our model mainly focus on brain-to-text decoding, improving SOTA from 6.94 to 12.92 on BLEU-1 and 6.93 to 13.06 on ROUGE-1F. It can also simulate brain signals, thereby serving as a novel neural interface. Code is available at \href{https://github.com/NeuSpeech/NeuGPT}{NeuSpeech/NeuGPT (https://github.com/NeuSpeech/NeuGPT) .}