AM
Avi Mendelson
cs.LGcs.CVcs.AIcs.ARcs.CLcs.CRcs.DCcs.NEcs.PFstat.ML
On Valency
published · living versionsW_7vj6bpzm·v1 · currentpublished
The Geometry of Ignorance: Detecting Suppressed Knowledge in LLMs through Activation Fragmentation
with Adar Peleg, Dvir Alsheich, Tomer Ashuach
1 version
W_26qwtbsy·v1published
Preprints & journals
37 papers in the corpus · 2018–2026Different Tasks, Different Timesteps: Temporal Task-Aware Quantization for Diffusion Language Models2610.31198v1 · Ofek Avivi, Eliron Rahimi, Avi Mendelson et al.2026 · 0 citationsarXivon Valency
On the SPOT: Enhancing Speculative Decoding for Diffusion Language Models via Commit-Order Prediction2610.22543v2 · Eliezer Shlomi, Eliron Rahimi, Avi Mendelson et al.2026 · 0 citationsarXiv
Generalization Is Stability, Not Accuracy: Multi-Axis Evaluation of LLMs2610.01428v1 · Nagham Omar, Mahmoud Jabarin, Maya Rozenshtein et al.2026 · 0 citationsarXiv
AnchorRep: Defending LLMs Against Cross-Model Adversarial Transfer via Representation Repulsion2609.32602v1 · Gal Wertheizer, Rom Himelstein, Tomer Peretz et al.2026 · 0 citationsarXiv
LGNNIC: Acceleration of Large-Scale GNN Training using SmartNICs2608.07733v1 · Liad Gerstman, Aditya Dhakal, Dejan Milojicic et al.2026 · 0 citationsarXiv
Linguistic Firewall: Geometry as Defense in Multi-Agent Systems Routing2606.30555v3 · Dvir Alsheich, Adar Peleg, Ben Hagag et al.2026 · 0 citationsarXiv
Breaking the Bubble: Asynchronous Pipeline Parallel Training with Bounded Weight Inconsistency2606.07881v1 · Itay Elam, Eliron Rahimi, Avi Mendelson et al.2026 · 0 citationsarXiv
Domain Restriction via Multi SAE Layer Transitions2605.11920v1 · Elias Shaheen, Avi Mendelson2026 · 0 citationsarXiv
Extracting Recurring Vulnerabilities from Black-Box LLM-Generated Software2602.04894v4 · Tomer Kordonsky, Amit LeVi, Maayan Yamin et al.2026 · 0 citationsarXiv
Step-Wise Refusal Dynamics in Autoregressive and Diffusion Language Models2602.02600v4 · Eliron Rahimi, Elad Hirshel, Rom Himelstein et al.2026 · 0 citationsarXiv
REMIND: Input Loss Landscapes Reveal Residual Memorization in Post-Unlearning LLMs2511.04228v1 · Liran Cohen, Yaniv Nemcovesky, Avi Mendelson2025 · 0 citationsarXiv
: Reconstruction, Raw, and Rain: Deraining Directly in the Bayer Domain2509.24022v3 · Nate Rothschild, Moshe Kimhi, Avi Mendelson et al.2025 · 0 citationsarXiv
Representing LLMs in Prompt Semantic Task Space2509.22506v1 · Idan Kashani, Avi Mendelson, Yaniv Nemcovsky2025 · 0 citationsarXiv
Silent Tokens, Loud Effects: Padding in LLMs2510.01238v2 · Rom Himelstein, Amit LeVi, Yonatan Belinkov et al.2025 · 0 citationsarXiv
Benchmarking Adversarial Patch Selection and Location2508.01676v1 · Shai Kimhi, Avi Mendlson, Moshe Kimhi2025 · 1 citationarXiv
Jailbreak Attack Initializations as Extractors of Compliance Directions2502.09755v4 · Amit Levi, Rom Himelstein, Yaniv Nemcovsky et al.2025 · 0 citationsarXiv
Sparse patches adversarial attacks via extrapolating point-wise information2411.16162v1 · Yaniv Nemcovsky, Avi Mendelson, Chaim Baskin2024 · 0 citationsarXiv
Hysteresis Activation Function for Efficient Inference2411.10573v3 · Moshe Kimhi, Idan Kashani, Avi Mendelson et al.2024 · 2 citationsProceedings of Machine Learning Research, Volume 262, Pages 414 422, 2024
SCART: Simulation of Cyber Attacks for Real-Time2304.03657v2 · Eliron Rahimi, Kfir Girstein, Roman Malits et al.2023 · 0 citationsProc. 15th Int. Conf. SIMULTECH, SciTePress, 2025, pp. 27-38
AMED: Automatic Mixed-Precision Quantization for Edge Devices2205.15437v2 · Moshe Kimhi, Tal Rozen, Avi Mendelson et al.2022 · 10 citationsarXiv
Bimodal Distributed Binarized Neural Networks2204.02004v1 · Tal Rozen, Moshe Kimhi, Brian Chmiel et al.2022 · 3 citationsarXiv
Weisfeiler and Leman Go Infinite: Spectral and Combinatorial Pre-Colorings2201.13410v2 · Or Feldman, Amit Boyarski, Shai Feldman et al.2022 · 1 citationarXiv
Graph Representation Learning via Aggregation Enhancement2201.12843v4 · Maxim Fishman, Chaim Baskin, Evgenii Zheltonozhskii et al.2022 · 0 citationsarXiv
Contrast to Divide: Self-Supervised Pre-Training for Learning with Noisy Labels2103.13646v2 · Evgenii Zheltonozhskii, Chaim Baskin, Avi Mendelson et al.2021 · 100 citationsarXiv
FlexWatts: A Power- and Workload-Aware Hybrid Power Delivery Network for Energy-Efficient Microprocessors2009.09094v1 · Jawad Haj-Yahya, Mohammed Alser, Jeremie S. Kim et al.2020 · 12 citationsarXiv
Asymmetric Aging Effect on Modern Microprocessors2009.03945v1 · Freddy Gabbay, Avi Mendelson2020 · 15 citationsarXiv
Self-Supervised Learning for Large-Scale Unsupervised Image Clustering2008.10312v2 · Evgenii Zheltonozhskii, Chaim Baskin, Alex M. Bronstein et al.2020 · 4 citationsarXiv
Electromigration-Aware Architecture for Modern Microprocessors2005.01593v3 · Freddy Gabbay, Avi Mendelson2020 · 4 citationsarXiv
HCM: Hardware-Aware Complexity Metric for Neural Network Architectures2004.08906v2 · Alex Karbachevsky, Chaim Baskin, Evgenii Zheltonozhskii et al.2020 · 19 citationsarXiv
Colored Noise Injection for Training Adversarially Robust Neural Networks2003.02188v2 · Evgenii Zheltonozhskii, Chaim Baskin, Yaniv Nemcovsky et al.2020 · 1 citationarXiv
Loss Aware Post-training Quantization1911.07190v2 · Yury Nahshan, Brian Chmiel, Chaim Baskin et al.2019 · 154 citationsarXiv
CAT: Compression-Aware Training for bandwidth reduction1909.11481v1 · Chaim Baskin, Brian Chmiel, Evgenii Zheltonozhskii et al.2019 · 11 citationsarXiv
Feature Map Transform Coding for Energy-Efficient CNN Inference1905.10830v3 · Brian Chmiel, Chaim Baskin, Ron Banner et al.2019 · 24 citationsarXiv
Towards Learning of Filter-Level Heterogeneous Compression of Convolutional Neural Networks1904.09872v4 · Yochai Zur, Chaim Baskin, Evgenii Zheltonozhskii et al.2019 · 5 citationsarXiv
Efficient non-uniform quantizer for quantized neural network targeting reconfigurable hardware1811.10869v1 · Natan Liss, Chaim Baskin, Avi Mendelson et al.2018 · 4 citationsarXiv
NICE: Noise Injection and Clamping Estimation for Neural Network Quantization1810.00162v2 · Chaim Baskin, Natan Liss, Yoav Chai et al.2018 · 34 citationsarXiv
ScaleSimulator: A Fast and Cycle-Accurate Parallel Simulator for Architectural Exploration1803.11440v1 · Ori Chalak, Cai Weiguang, Li Wei et al.2018 · 0 citationsarXiv
Career total: 172 works. 37 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.