SA
Seyedarmin Azizi
cs.AIcs.LGstat.MLcs.CLcs.CVcs.NEeess.IVstat.AP
On Valency
published · living versionsW_wwmttknn·v1 · currentpublished
When Pointing Becomes Profiling: Social Bias and Refusal in Text-Steerable Visual Encoders
with Parsa Razmara, Sajjad Ghiasvand, Ardalan Aryashad, Tina Khezresmaeilzadeh
1 version
W_fqedquhu·v1published
W_nmut63vy·v1published
W_28b37zwp·v1published
Preprints & journals
23 papers in the corpus · 2023–2026Keep the Lineages Alive: Diversity-Preserving SMC for Diffusion LLM Decoding2610.25932v1 · Minoo Ahmadi, Seyedarmin Azizi, Mehdi Kamal et al.2026 · 0 citationsarXivon Valency
Claim Laundering: When Agent Decision Models Trust Records Over Reality2610.25947v1 · Tina Khezresmaeilzadeh, Seyedarmin Azizi, Erfan Baghaei Potraghloo et al.2026 · 0 citationsarXivon Valency
One Skill Takes Over: Clashing Agent Skills and a One-File Fix2610.25064v1 · Parsa Razmara, Seyedarmin Azizi, Ardalan Aryashad2026 · 0 citationsarXivon Valency
The Sharp Directions Are Against You: Curvature Analysis of Activation Steering2610.16084v1 · Seyedarmin Azizi, Arya Fayyazi, Parsa Razmara et al.2026 · 0 citationsarXiv
Running Fewer Loops Removes Refusal Behaviour in Recurrent-Depth Language Models2610.15218v1 · Seyedarmin Azizi, Erfan Baghaei Potraghloo, Arya Fayyazi et al.2026 · 0 citationsarXiv
One Word Opens the Gate: The Option-Channel Attack on Typed Decision Models as Agent Guardrails2610.12292v1 · Seyedarmin Azizi, Erfan Baghaei Potraghloo, Massoud Pedram2026 · 0 citationsarXiv
Sharpen Without Search: On-Policy Distillation of Sequence-Level Power Distribution2610.06804v1 · Erfan Baghaei Potraghloo, Seyedarmin Azizi, Arya Fayyazi et al.2026 · 0 citationsarXiv
Labels Override Definitions in Jev-Style Typed Decision Models2610.02586v1 · Seyedarmin Azizi, Erfan Baghaei Potraghloo, Massoud Pedram2026 · 0 citationsarXiv
Chopthin-Consensus Power Sampling: A Diversity-Preserving Approach to LLM Decoding2609.12243v2 · Minoo Ahmadi, Seyedarmin Azizi, Erfan Baghaei Potraghloo et al.2026 · 0 citationsarXiv
CircuitSteer: Geometrically Aligned Multi-Layer Steering via Sparse Autoencoder Circuits2608.05732v1 · Mehrshad Saadatinia, Parsa Razmara, Ardalan Aryashad et al.2026 · 0 citationsarXiv
One Token Away from Collapse: The Fragility of Instruction-Tuned Helpfulness2604.13006v2 · Erfan Baghaei Potraghloo, Seyedarmin Azizi, Souvik Kundu et al.2026 · 0 citationsarXiv
Power-SMC: Low-Latency Sequence-Level Power Sampling for Training-Free LLM Reasoning2602.10273v3 · Seyedarmin Azizi, Erfan Baghaei Potraghloo, Minoo Ahmadi et al.2026 · 0 citationsarXiv
From Filters to VLMs: Benchmarking Defogging Methods through Object Detection and Segmentation Performance2510.03906v2 · Ardalan Aryashad, Parsa Razmara, Amin Mahjoub et al.2025 · 0 citationsarXiv
Top-H Decoding: Adapting the Creativity and Coherence with Bounded Entropy in Text Generation2509.02510v2 · Erfan Baghaei Potraghloo, Seyedarmin Azizi, Souvik Kundu et al.2025 · 1 citationarXiv
Activation Steering for Chain-of-Thought Compression2507.04742v2 · Seyedarmin Azizi, Erfan Baghaei Potraghloo, Massoud Pedram2025 · 0 citationsarXiv
VISTA: Vision-Language Inference for Training-Free Stock Time-Series Analysis2505.18570v4 · Tina Khezresmaeilzadeh, Parsa Razmara, Seyedarmin Azizi et al.2025 · 0 citationsarXiv
Efficient Noise Mitigation for Enhancing Inference Accuracy in DNNs on Mixed-Signal Accelerators2409.18553v1 · Seyedarmin Azizi, Mohammad Erfan Sadeghi, Mehdi Kamal et al.2024 · 0 citationsarXiv
PEANO-ViT: Power-Efficient Approximations of Non-Linearities in Vision Transformers2406.14854v2 · Mohammad Erfan Sadeghi, Arash Fayyazi, Seyedarmin Azizi et al.2024 · 17 citationsarXiv
LaMDA: Large Model Fine-Tuning via Spectrally Decomposed Low-Dimensional Adaptation2406.12832v1 · Seyedarmin Azizi, Souvik Kundu, Massoud Pedram2024 · 5 citationsarXiv
Memory-Efficient Vision Transformers: An Activation-Aware Mixed-Rank Compression Strategy2402.06004v1 · Seyedarmin Azizi, Mahdi Nazemi, Massoud Pedram2024 · 1 citationarXiv
Low-Precision Mixed-Computation Models for Inference on Edge2312.02210v1 · Seyedarmin Azizi, Mahdi Nazemi, Mehdi Kamal et al.2023 · 7 citationsarXiv
Sensitivity-Aware Mixed-Precision Quantization and Width Optimization of Deep Neural Networks Through Cluster-Based Tree-Structured Parzen Estimation2308.06422v3 · Seyedarmin Azizi, Mahdi Nazemi, Arash Fayyazi et al.2023 · 2 citationsarXiv
CrAFT: Compression-Aware Fine-Tuning for Efficient Visual Task Adaptation2305.04526v2 · Jung Hwan Heo, Seyedarmin Azizi, Arash Fayyazi et al.2023 · 1 citationarXiv
Career total: 28 works. 23 are in this corpus.
Profile built from the corpus for this byline.
Author records are still filling in while the Hub is in alpha. If this is your page, you'll be able to claim it soon. Spot a mistake? Tell us.