The SAIRC Journal

Full credit goes to the original author, linked below. All research papers were reposted either with permission of the author, or by anonymous submission by SAIRC members like yourself. Large Language Models (LLMs) often generate inconsistent responses when prompted with semantically equivalent paraphrased inputs. Recently, activation steering, a technique that modulates LLMs' behaviours by adjusting their latent representations during inference time, has been explored to improve the semantic consistency of LLMs. However, these methods typically operate at the model component level, such as layer hidden states or attention head outputs. They face a challenge due to the ``polysemanticity issue'', where the model components of LLMs typically encode multiple entangled features, making precise steering difficult. To address this challenge, we drill down to feature-level representations and propose LF-Steering, a novel activation steering approach to precisely identify latent feature representations responsible for semantic inconsistency. More specifically, our method maps the hidden states of the relevant transformer layer into a sparsely activated, high-dimensional feature space based on a sparse autoencoder (SAE), ensuring model steering based on decoupled feature representations with minimal interference. Comprehensive experiments on NLU and NLG datasets demonstrate the effectiveness of our method in enhancing semantic consistency, resulting in significant performance gains for various NLU and NLG tasks.

Full credit goes to the original author, linked below. All research papers were reposted either with permission of the author, or by anonymous submission by SAIRC members like yourself. Effective and reliable control over large language model (LLM) behavior is a significant challenge. While activation steering methods, which add steering vectors to a model's hidden states, are a promising approach, existing techniques often lack precision and interpretability in how they influence model outputs. We introduce Feature Guided Activation Additions (FGAA), a novel activation steering method that leverages insights from Contrastive Activation Addition (CAA) and Sparse Autoencoder-Targeted Steering (SAE-TS). By operating in the latent space of a Sparse Autoencoder (SAE) and employing optimization techniques to select desired SAE features, FGAA constructs precise steering vectors that provide better steering effects while maintaining coherence of steered model outputs. In this regard, evaluations on Gemma-2-2B and Gemma-2-9B models across various steering tasks demonstrate that FGAA outperforms existing steering methods of CAA, SAE decoder steering, and SAE-TS. Our results also highlight important trade-offs between steering scale and general model capabilities that are consistent across all tested steering methods.

In the remote regions of Northern Tanzania, women and children of the Maasai Tribe walk nine hours a day to collect water for their families. Over four years, the collaborative efforts with the Maasai communities have led to the installation of four water harvesting units, enhancing the local socio-economic conditions by facilitating educational opportunities and economic pursuits for over 4,500 individuals within a 10-mile radius. This project presents a novel approach to addressing this issue by integrating satellite data and image classification to identify densely populated areas marked by uniquely shaped Maasai homes lacking a water supply and planning the best placement of rainwater harvesting units. The backbone of this project was developing an image classification model trained on 10,000 hand-selected satellite image samples of Bomas. This model generated a density heat map, enabling the strategic placement of water harvesting units in the most critical locations to maximize impact. Our findings underscore the potential of satellite technology in humanitarian interventions, particularly in harder-to-reach areas where traditional surveying and data collection techniques are impractical.

Full credit goes to the original author, linked below. All research papers were reposted either with permission of the author, or by anonymous submission by SAIRC members like yourself. Large Language Models (LLMs) have exhibited an impressive capability to perform reasoning tasks, especially if they are encouraged to generate a sequence of intermediate steps. Reasoning performance can be improved by suitably combining multiple LLM responses, generated either in parallel in a single query, or via sequential interactions with LLMs throughout the reasoning process. Existing strategies for combination, such as self-consistency and progressive-hint-prompting, make inefficient usage of the LLM responses. We present Refined Answer Distributions, a novel and principled algorithmic framework to enhance the reasoning capabilities of LLMs. Our approach can be viewed as an iterative sampling strategy for forming a Monte Carlo approximation of an underlying distribution of answers, with the goal of identifying the mode -- the most likely answer. Empirical evaluation on several reasoning benchmarks demonstrates the superiority of the proposed approach.

Full credit goes to the original author, linked below. All research papers were reposted either with permission of the author, or by anonymous submission by SAIRC members like yourself. The rapid development of deep learning has significantly improved salient object detection (SOD) combining both RGB and thermal (RGB-T) images. However, existing Transformer-based RGB-T SOD models with quadratic complexity are memory-intensive, limiting their application in high-resolution bimodal feature fusion. To overcome this limitation, we propose a purely Fourier Transform-based model, namely Deep Fourier-embedded Network (FreqSal), for accurate RGB-T SOD. Specifically, we leverage the efficiency of Fast Fourier Transform with linear complexity to design three key components: (1) To fuse RGB and thermal modalities, we propose Modal-coordinated Perception Attention, which aligns and enhances bimodal Fourier representation in multiple dimensions; (2) To clarify object edges and suppress noise, we design Frequency-decomposed Edge-aware Block, which deeply decomposes and filters Fourier components of low-level features; (3) To accurately decode features, we propose Fourier Residual Channel Attention Block, which prioritizes high-frequency information while aligning channel-wise global relationships. Additionally, even when converged, existing deep learning-based SOD models' predictions still exhibit frequency gaps relative to ground-truth. To address this problem, we propose Co-focus Frequency Loss, which dynamically weights hard frequencies during edge frequency reconstruction by cross-referencing bimodal edge information in the Fourier domain. Extensive experiments on ten bimodal SOD benchmark datasets demonstrate that FreqSal outperforms twenty-nine existing state-of-the-art bimodal SOD models. Comprehensive ablation studies further validate the value and effectiveness of our newly proposed components. The code is available at https://github.com/JoshuaLPF/FreqSal.

Full credit goes to the original author, linked below. All research papers were reposted either with permission of the author, or by anonymous submission by SAIRC members like yourself. Responsible deployment of language models requires mechanisms for refusing unsafe prompts while preserving model performance. While most approaches modify model weights through additional training, we explore an alternative: steering model activations at inference time via amplifying sparse autoencoder (SAE) features that mediate refusal. This work uncovers a fundamental tension between SAE steering-based safety improvements and general model capabilities. While feature steering successfully improves robustness against both single-turn and challenging multi-turn jailbreak attempts, we discover that this comes at a previously underexplored cost -- systematic degradation of performance across multiple benchmark tasks, even on safe inputs with no apparent connection to refusal behavior. This suggests that features mediating refusal may be more deeply entangled with general language model capabilities than previously understood. Our findings reveal important open questions about the nature of safety-relevant features in language models and the feasibility of isolating them for targeted intervention. While SAE-based steering shows promise as a flexible approach to enhancing language model safety, our results highlight the critical need to understand and address the mechanisms behind these capability tradeoffs before such techniques can be practically deployed.

Reproduced with permission of author. Mitchell examines how the metaphors we use to describe AI systems—from "neural networks" to "hallucinations" to "understanding"—shape our expectations and interpretations of these technologies. Drawing on Terrence Sejnowski's reflection that a threshold was crossed when LLMs began communicating "in an eerily human way," the piece argues that better metaphors are needed to accurately characterize what AI systems are and, crucially, what they are not.

Full credit goes to the original author, linked below. All research papers were reposted either with permission of the author, or by anonymous submission by SAIRC members like yourself. To control the behavior of language models, steering methods attempt to ensure that outputs of the model satisfy specific pre-defined properties. Adding steering vectors to the model is a promising method of model control that is easier than finetuning, and may be more robust than prompting. However, it can be difficult to anticipate the effects of steering vectors produced by methods such as CAA [Panickssery et al., 2024] or the direct use of SAE latents [Templeton et al., 2024]. In our work, we address this issue by using SAEs to measure the effects of steering vectors, giving us a method that can be used to understand the causal effect of any steering vector intervention. We use this method for measuring causal effects to develop an improved steering method, SAE-Targeted Steering (SAE-TS), which finds steering vectors to target specific SAE features while minimizing unintended side effects. We show that overall, SAE-TS balances steering effects with coherence better than CAA and SAE feature steering, when evaluated on a range of tasks.

Multimodal contrastive learning trains neural networks by leveraging data from heterogeneous sources such as images and text. Yet, many current multimodal learning architectures cannot generalize to an arbitrary number of modalities and need to be hand-constructed. We propose AutoBIND, a novel contrastive learning framework that can learn representations from an arbitrary number of modalities through graph optimization. We evaluate AutoBIND on Alzheimer's disease detection because it has real-world medical applicability and it contains a broad range of data modalities. We show that AutoBIND outperforms previous methods on this task, highlighting the generalizability of the approach. NOTE: Anyone interested should also consider reading this ICLR blog post for context: https://iclr-blogposts.github.io/2025/blog/multimodal-learning/

Reproduced for educational purposes with permission from Xuefeng et. al. Vision-language models (VLMs) are essential for contextual understanding of both visual and textual information. However, their vulnerability to adversarially manipulated inputs presents significant risks, leading to compromised outputs and raising concerns about the reliability in VLM-integrated applications. Detecting these malicious prompts is thus crucial for maintaining trust in VLM generations. A major challenge in developing a safeguarding prompt classifier is the lack of a large amount of labeled benign and malicious data. To address the issue, we introduce VLMGuard, a novel learning framework that leverages the unlabeled user prompts in the wild for malicious prompt detection. These unlabeled prompts, which naturally arise when VLMs are deployed in the open world, consist of both benign and malicious information. To harness the unlabeled data, we present an automated maliciousness estimation score for distinguishing between benign and malicious samples within this unlabeled mixture, thereby enabling the training of a binary prompt classifier on top. Notably, our framework does not require extra human annotations, offering strong flexibility and practicality for real-world applications. Extensive experiment shows VLMGuard achieves superior detection results, significantly outperforming state-of-the-art methods.

"Can machines think?" So asked Alan Turing in his 1950 paper, "Computing Machinery and Intelligence." Turing quickly noted that, given the difficulty of defining thinking, the question is "too meaningless to deserve discussion." As is often done in philosophical debates, he proposed replacing it with a different question. Turing imagined an "imitation game," in which a human judge converses with both a computer and a human (a "foil"), each of which vies to convince the judge that they are the human. Importantly, the computer, foil, and judge do not see one another; they communicate entirely through text. After conversing with each candidate, the judge guesses which one is the real human. Turing's new question was, "Are there imaginable digital computers which would do well in the imitation game?"

Full credit goes to the original author, linked below. All research papers were reposted either with permission of the author, or by anonymous submission by SAIRC members like yourself. Mechanistic interpretability (MI) is an emerging sub-field of interpretability that seeks to understand a neural network model by reverse-engineering its internal computations. Recently, MI has garnered significant attention for interpreting transformer-based language models (LMs), resulting in many novel insights yet introducing new challenges. However, there has not been work that comprehensively reviews these insights and challenges, particularly as a guide for newcomers to this field. To fill this gap, we provide a comprehensive survey from a task-centric perspective, organizing the taxonomy of MI research around specific research questions or tasks. We outline the fundamental objects of study in MI, along with the techniques, evaluation methods, and key findings for each task in the taxonomy. In particular, we present a task-centric taxonomy as a roadmap for beginners to navigate the field by helping them quickly identify impactful problems in which they are most interested and leverage MI for their benefit. Finally, we discuss the current gaps in the field and suggest potential future directions for MI research.

The term "artificial general intelligence" (AGI) has become ubiquitous in current discourse around AI. OpenAI states that its mission is "to ensure that artificial general intelligence benefits all of humanity." DeepMind's company vision statement notes that "artificial general intelligence…has the potential to drive one of the greatest transformations in history." AGI is mentioned prominently in the UK government's National AI Strategy and in US government AI documents. Microsoft researchers recently claimed evidence of "sparks of AGI" in the large language model GPT-4, and current and former Google executives proclaimed that "AGI is already here." The question of whether GPT-4 is an "AGI algorithm" is at the center of a lawsuit filed by Elon Musk against OpenAI.

Reproduced with permission of author. Mitchell explores the fundamental challenge of getting AI systems to understand the world as humans do, illustrating the gap through real-world examples of AI failures in deployed contexts—including vision systems that misidentify billboard stop signs as real traffic signals. The piece argues that current AI systems, despite impressive capabilities, lack the grounded, contextual understanding needed for robust real-world deployment, and examines what bridging that gap would require.

Reproduced with permission of author. Revisiting Marvin Minsky's 1967 prediction that AI would be "substantially solved" within a generation, Mitchell asks how we should measure and evaluate progress toward human-level machine intelligence nearly 60 years later. The piece questions current AI benchmarking methodologies and argues that evaluating machine intelligence requires a deeper reckoning with what we mean by "intelligence" itself.