arXiv subjects

Ali Ebrahimpour-Boroojeny

Publications and source records attributed to Ali Ebrahimpour-Boroojeny.

9 recordsLinked to original sources

arXiv2026

Resisting Harmful Instructions through Training-Free Weight Editing

Instruction-following language models can comply with requests that conflict with intended behaviors and safety alignment. We introduce a training-free weight-editing approach for reducing compliance with targeted undesirable instructions. Our method needs only forward passes on a small set of prompt pairs per target behavior, and because the edit is stored in the model's weights, changing or deleting instructions in the prompt cannot undo the weight modification. We evaluate the approach on two target harmful behaviors across five instruction-tuned models: instructed falsehood and instructed sexism. On two instructed-lying benchmarks, it improves mean honest response rates by 6.2 and 4.6 percentage points over the strongest alternative, significantly on three of the five models and within noise on the other two. In a controlled study of seven prompt formats, it maintains higher honest-response rates than single-layer activation steering when orders to lie arrive in single-turn formats not used to construct the edit. For instructed sexism, it reduces the mean sexist-response rate from 57.9% to 24.3%; combining the edit with a fairness prompt reduces it further to 11.9%. These results establish our contrast-based weight editing approach as an effective training-free intervention in two distinct behavioral domains, supporting its use both independently and alongside prompt-based safeguards.

cs.LG↗

arXiv2026

You Can't Have It Both Ways: Concept Entanglement Limits Diffusion Model Unlearning

Concept unlearning in text-to-image diffusion models aims to suppress a target concept (e.g., \texttt{horse}) while preserving related but distinct content (e.g., \texttt{donkey}), yet existing methods either leak under indirect prompts or visibly degrade other concepts. We show that these failure modes stem from the geometry of concept representations rather than from any particular algorithm. Formalizing concepts as activation-space regions, we prove that the overlap between a target and other concepts lower-bounds the damage any robust erasure must inflict on them, with the trade-off scaling linearly in the degree of overlap. Across thirteen unlearning methods, including methods designed to preserve non-target concepts, no method achieves both strong erasure and strong neighbor preservation: STEREO nearly eliminates indirect leakage but cuts neighbor generation by more than 75\%, while sparse inference-time methods preserve neighbors but leak. Damage increases with our overlap measure, monotonically so for STEREO; the $κ$-scaling reproduces on SDXL, and neighbor-selective damage recurs on FLUX. Perfect unlearning is the wrong target for entangled concepts; methods should be evaluated on the Pareto frontier our theorem establishes.

cs.AI↗

arXiv2026

On the Necessity of Output Distribution Reweighting for Effective Class Unlearning

In this paper, we reveal a significant shortcoming in class unlearning evaluations: overlooking the underlying class geometry can cause information leakage about the forgotten class. We further propose a simple unlearning strategy to mitigate this issue. We introduce Class Membership Inference Attack (CMIA) that uses the probabilities assigned by the model to neighboring classes to detect unlearned samples. We find that existing unlearning methods are vulnerable to CMIA across multiple datasets. We then propose a new fine-tuning objective that mitigates this privacy leakage by approximating, for forget-class inputs, the distribution over the remaining classes that a retrained-from-scratch model would produce. To construct this approximation, we estimate inter-class similarity and tilt the target model's distribution accordingly. The resulting Tilted REWeighting (TREW) distribution serves as the desired distribution during fine-tuning. We also show that across multiple benchmarks, TREW matches or surpasses existing unlearning methods on prior unlearning metrics. More specifically, on CIFAR-10, it reduces the gap with retrained models by 19% and 46% for U-LiRA and CMIA scores, accordingly, compared to the SOTA method for each category. Our code is publicly available at https://github.com/CrowdDynamicsLab/ICML_TREW.

cs.LG↗

arXiv2026

AgentAbstain: Do LLM Agents Know When Not to Act?

Agent systems based on large language models (LLMs) are increasingly deployed for autonomous tasks, yet existing evaluations mostly focus on task success rather than whether agents know when to abstain. This gap poses real risks: under ambiguity, conflicting constraints, or tool failures, agents may execute unintended and irreversible actions. To close this gap, we present the first systematic evaluation framework for agentic abstention: the calibrated ability of tool-using LLM agents to recognize when not to act. At its core, AgentAbstain is a paired-task benchmark built on an agent-native taxonomy of 8 abstention scenarios across pre-execution reasoning and runtime discovery. It contains 263 paired tasks across 42 executable sandbox environments, where each pair consists of a should-act task and a should-abstain variant produced through a controlled perturbation to the instruction, tool, or environment state. To scale this paired design and resist data contamination, we propose AbstainGen, a fully automated pipeline that synthesizes sandbox environments and generates paired tasks end-to-end, validated by deterministic replay and semantic LLM judges; fresh task instances can be regenerated on demand, and three independent annotators rate 94-98% of sampled tasks as well-designed. Across 17 frontier LLMs in 4 agent harnesses, the best agent (Gemini 3.1 Pro) achieves only 59.5% paired accuracy (correct on both the act and abstain sides of each paired task). More importantly, abstention capability is largely independent of general task-solving capability, indicating that scaling task-solving alone will not close this gap. We further identify failure modes such as post-hoc abstention, in which agents execute irreversible actions before recognizing abstention triggers. Our code and dataset are open-sourced at agentabstain.github.io.

cs.AI↗

arXiv2025

Toward Reliable Machine Unlearning: Theory, Algorithms, and Evaluation

We propose new methodologies for both unlearning random set of samples and class unlearning and show that they outperform existing methods. The main driver of our unlearning methods is the similarity of predictions to a retrained model on both the forget and remain samples. We introduce Adversarial Machine UNlearning (AMUN), which surpasses prior state-of-the-art methods for image classification based on SOTA MIA scores. AMUN lowers the model's confidence on forget samples by fine-tuning on their corresponding adversarial examples. Through theoretical analysis, we identify factors governing AMUN's performance, including smoothness. To facilitate training of smooth models with a controlled Lipschitz constant, we propose FastClip, a scalable method that performs layer-wise spectral-norm clipping of affine layers. In a separate study, we show that increased smoothness naturally improves adversarial example transfer, thereby supporting the second factor above. Following the same principles for class unlearning, we show that existing methods fail in replicating a retrained model's behavior by introducing a nearest-neighbor membership inference attack (MIA-NN) that uses the probabilities assigned to neighboring classes to detect unlearned samples and demonstrate the vulnerability of such methods. We then propose a fine-tuning objective that mitigates this leakage by approximating, for forget-class inputs, the distribution over remaining classes that a model retrained from scratch would produce. To construct this approximation, we estimate inter-class similarity and tilt the target model's distribution accordingly. The resulting Tilted ReWeighting(TRW) distribution serves as the desired target during fine-tuning. Across multiple benchmarks, TRW matches or surpasses existing unlearning methods on prior metrics.

cs.LG↗

arXiv2025

AMUN: Adversarial Machine UNlearning

Machine unlearning, where users can request the deletion of a forget dataset, is becoming increasingly important because of numerous privacy regulations. Initial works on ``exact'' unlearning (e.g., retraining) incur large computational overheads. However, while computationally inexpensive, ``approximate'' methods have fallen short of reaching the effectiveness of exact unlearning: models produced fail to obtain comparable accuracy and prediction confidence on both the forget and test (i.e., unseen) dataset. Exploiting this observation, we propose a new unlearning method, Adversarial Machine UNlearning (AMUN), that outperforms prior state-of-the-art (SOTA) methods for image classification. AMUN lowers the confidence of the model on the forget samples by fine-tuning the model on their corresponding adversarial examples. Adversarial examples naturally belong to the distribution imposed by the model on the input space; fine-tuning the model on the adversarial examples closest to the corresponding forget samples (a) localizes the changes to the decision boundary of the model around each forget sample and (b) avoids drastic changes to the global behavior of the model, thereby preserving the model's accuracy on test samples. Using AMUN for unlearning a random $10\%$ of CIFAR-10 samples, we observe that even SOTA membership inference attacks cannot do better than random guessing.

cs.LG↗

arXiv2025

Small Cues, Big Differences: Evaluating Interaction and Presentation for Annotation Retrieval in AR

Augmented Reality (AR) enables intuitive interaction with virtual annotations overlaid on the real world, supporting a wide range of applications such as remote assistance, education, and industrial training. However, as the number of heterogeneous annotations increases, their efficient retrieval remains an open challenge in 3D environments. This paper examines how interaction modalities and presentation designs affect user performance, workload, fatigue, and preference in AR annotation retrieval. In two user studies, we compare eye-gaze versus hand-ray hovering and evaluate four presentation methods: Opacity-based, Scale-based, Nothing-based, and Marker-based. Results show that eye-gaze was favored over hand-ray by users, despite leading to significantly higher unintentional activations. Among the presentation methods, Scale-based presentation reduces workload and task completion time while aligning with user preferences. Our findings offer empirical insights into the effectiveness of different annotation presentation methods, leading to design recommendations for building more efficient and user-friendly AR annotation review systems.

cs.HC↗

arXiv2024

LOTOS: Layer-wise Orthogonalization for Training Robust Ensembles

Transferability of adversarial examples is a well-known property that endangers all classification models, even those that are only accessible through black-box queries. Prior work has shown that an ensemble of models is more resilient to transferability: the probability that an adversarial example is effective against most models of the ensemble is low. Thus, most ongoing research focuses on improving ensemble diversity. Another line of prior work has shown that Lipschitz continuity of the models can make models more robust since it limits how a model's output changes with small input perturbations. In this paper, we study the effect of Lipschitz continuity on transferability rates. We show that although a lower Lipschitz constant increases the robustness of a single model, it is not as beneficial in training robust ensembles as it increases the transferability rate of adversarial examples across models in the ensemble. Therefore, we introduce LOTOS, a new training paradigm for ensembles, which counteracts this adverse effect. It does so by promoting orthogonality among the top-$k$ sub-spaces of the transformations of the corresponding affine layers of any pair of models in the ensemble. We theoretically show that $k$ does not need to be large for convolutional layers, which makes the computational overhead negligible. Through various experiments, we show LOTOS increases the robust accuracy of ensembles of ResNet-18 models by $6$ percentage points (p.p) against black-box attacks on CIFAR-10. It is also capable of combining with the robustness of prior state-of-the-art methods for training robust ensembles to enhance their robust accuracy by $10.7$ p.p.

cs.LG↗

arXiv2020

BPPart and BPMax: RNA-RNA Interaction Partition Function and Structure Prediction for the Base Pair Counting Model

RNA-RNA interaction (RRI) is ubiquitous and has complex roles in the cellular functions. In human health studies, miRNA-target and lncRNAs are among an elite class of RRIs that have been extensively studied. Bacterial ncRNA-target and RNA interference are other classes of RRIs that have received significant attention. In recent studies, mRNA-mRNA interaction instances have been observed, where both partners appear in the same pathway without any direct link between them, or any prior knowledge about their relationship. Those recently discovered cases suggest that RRI scope is much wider than those aforementioned elite classes. We revisit our RNA-RNA interaction partition function algorithm, piRNA, which computes the partition function, base-pairing probabilities, and structure for the comprehensive Turner energy model using 96 different dynamic programming tables. In this study, we strategically retreat from sophisticated thermodynamic models to the much simpler base pair counting model. That might seem counter-intuitive at the first glance; our idea is to benefit from the advantages of such simple models in terms of running time and memory footprint and compensate for the associated information loss by adding machine learning components in the future. Here, simple weighted base pair counting is considered to obtain BPPart for Base-pair Partition function and BPMax for Base-pair Maximization, which use 9 and 2 tables respectively. They are empirically 225 and 1350 fold faster than piRNA. A correlation of 0.855 and 0.836 was achieved between piRNA and BPPart and between piRNA and BPMax, respectively, in 37 degrees, and 0.920 and 0.904 in -180 degrees. We also discover two partner RNAs, SNORD3D and TRAF3, and hypothesize their potential roles in genetic diseases. We envision fusion of machine learning methods with the proposed algorithms in the future.

q-bio.BM↗