{"total":16,"items":[{"citing_arxiv_id":"2606.27934","ref_index":10,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Self-Verifying Measurement Records: Hash-Linked Evidence Graphs for Hardware Benchmarking","primary_cat":"cs.CR","submitted_at":"2026-06-26T10:26:50+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"The paper constructs hash-linked evidence graphs that bind hardware measurement quantities to their verification records, enabling offline auditing with probabilistic matrix checks and security measures against probe attacks on GPUs.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.21791","ref_index":2,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Single-Event Upsets in 3D Gaussian Splatting Rendering: Bit-Level Criticality, Spatial Extent, and a Parallel Support Guard","primary_cat":"cs.GR","submitted_at":"2026-06-19T22:54:25+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Bit flips in 3D Gaussian splatting are highly concentrated in effect with certain high-order bits corrupting up to 75.7% of the frame, but a support guard reduces the worst footprint to 11.68% while preserving clean performance and improving quality under accumulated faults.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.10651","ref_index":90,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Kwai Keye-VL-2.0 Technical Report","primary_cat":"cs.CV","submitted_at":"2026-06-09T09:58:08+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":4.0,"formal_verification":"none","one_line_summary":"Kwai Keye-VL-2.0-30B-A3B is a 30B MoE model with 3B active parameters using DSA adaptation and MOPD distillation that reports SOTA results on video understanding and agent benchmarks.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.05433","ref_index":10,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Zero knowledge verification for frontier AI training is possible","primary_cat":"cs.AI","submitted_at":"2026-06-03T20:57:28+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Proposes zkVM-based protocol for verifiable frontier AI pre-training with committed specs, network observations, Merkle commitments, and FP precompiles, estimating 36-month POC at single-digit overhead.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.02430","ref_index":21,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Not All Errors Are Equal: A Systematic Study of Error Propagation in Large Language Model Inference","primary_cat":"cs.DC","submitted_at":"2026-06-01T16:04:51+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"A new fault-injection framework enables a systematic empirical study that produces 17 takeaways on error propagation in LLM inference and four software-only mitigation directions.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.15638","ref_index":16,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"ITHICA: Intra-Thread Instruction Checking Approach for Defect-Induced Silent Data Corruptions","primary_cat":"cs.AR","submitted_at":"2026-05-15T05:43:33+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"ITHICA generates functional tests via intra-thread instruction duplication and comparison, detecting 39% more defective servers than baseline methods on over 3000 real CPUs while revealing new defect behaviors.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.08594","ref_index":6,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"FLARE: One-Shot PE-Level Fault Localization in Systolic Arrays via Algebraic Test Vectors","primary_cat":"cs.AR","submitted_at":"2026-05-09T01:22:28+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"FLARE uses pairwise coprime test vectors to create unique divisibility signatures that localize faulty rows in systolic arrays with one test pass and over 98% probability for 256x256 INT16 arrays.","context_count":1,"top_context_role":"background","top_context_polarity":"background","context_text":"In2025 IEEE International Symposium on High Performance Computer Architecture (HPCA), pages 775-789. IEEE, 2025. [5] Peter H. Hochschild, Paul Turner, Jeffrey C. Mogul, Rama Gober, Parthasarathy Ranganathan, David E. Culler, and Amin Vahdat. Cores that don't count. In Proceedings of the Workshop on Hot Topics in Operating Systems (HotOS), pages 9-16. ACM, 2021. [6] Harish Dattatraya Dixit, Sneha Pendharkar, Matt Beadon, Chris Mason, Tejasvi Chakravarthy, Bharath Muthiah, and Sriram Sankar. Silent data corruptions at scale. InarXiv preprint arXiv:2102.11245, 2021. [7] Yi He, Mike Hutton, Steven Chan, Robert De Gruijl, Rama Govindaraju, Nishant Patil, and Yanjing Li. Understanding permanent hardware failures in deep learn-"},{"citing_arxiv_id":"2605.07417","ref_index":5,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Effective and Memory-Efficient Alternatives to ECC for Reliable Large-Scale DNNs","primary_cat":"cs.AR","submitted_at":"2026-05-08T08:11:15+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"MSET and CEP deliver higher reliability than SECDED ECC for CNNs and Vision Transformers with zero memory overhead and substantially lower area and delay.","context_count":1,"top_context_role":"background","top_context_polarity":"background","context_text":"In safety-critical EdgeAI applications, such as autonomous ve- hicles, hardware faults can result in severe or even catastrophic consequences [4]. Similarly, in High-Performance Computing (HPC) environments, a large number of faults manifest as Silent Data Corruptions (SDCs) on a daily basis, adversely affecting the reliable deployment of ML models on large-scale server infrastructures [5], [6]. These failure modes pose significant challenges to the correct- ness and robustness of ML-driven workloads. The scale of modern DNNs further necessitates the need for fault-tolerance mechanisms in memories. State-of-the-art DNNs commonly comprise hundreds of millions to billions of parameters [7], all of which are stored in DRAM and SRAM memories that are inherently susceptible to soft"},{"citing_arxiv_id":"2605.04213","ref_index":3,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"The Anatomy of Silent Data Corruption: GPU Error Pattern Study and Modeling Guidance","primary_cat":"cs.AR","submitted_at":"2026-05-05T18:54:19+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Large-scale GPU fault injection shows NaN/inf outcomes are only 1% of SDC, single-bit flips under 40%, and corruption addresses are periodic, supporting distribution-aware modeling.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.22032","ref_index":11,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Kernel Contracts: A Specification Language for ML Kernel Correctness Across Heterogeneous Silicon","primary_cat":"cs.LG","submitted_at":"2026-04-23T19:46:52+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"Kernel Contracts is a specification language that formalizes correctness requirements for ML kernels to ensure consistent results across heterogeneous silicon platforms.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.20587","ref_index":42,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Making TransactionIsolation Checking Practical","primary_cat":"cs.DB","submitted_at":"2026-04-22T14:03:52+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"Boomslang introduces a front-end/back-end pipeline with superpositions in its IR to enable general-purpose checking of arbitrary transaction isolation levels via SMT solving.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.10390","ref_index":2,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"LLM-PRISM: Characterizing Silent Data Corruption from Permanent GPU Faults in LLM Training","primary_cat":"cs.AR","submitted_at":"2026-04-12T00:35:08+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"LLMs resist low-frequency permanent GPU faults but certain datapaths and precision formats trigger catastrophic training divergence even at moderate fault rates.","context_count":1,"top_context_role":"background","top_context_polarity":"background","context_text":"driven the deployment of massive-scale AI infrastructure, with state-of-the-art training runs now executing on clusters with tens of thousands GPUs spanning weeks to months [1]. As these systems scale, the reliability of the underlying hardware becomes increasingly crucial. Silent data corruption (SDC), faults that lead to incorrect computations without triggering detectable errors [2]-[4], poses a serious threat to the cor- rectness, stability, and reproducibility of LLM training [5]. Even minor deviations during training can compound through billions of updates, potentially degrading model accuracy, convergence behavior, or downstream model accuracy and/or performance in subtle and costly ways. Most prior studies assessing the resilience of deep learning"},{"citing_arxiv_id":"2604.09994","ref_index":10,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Aging Aware Adaptive Voltage Scaling for Reliable and Efficient AI Accelerators","primary_cat":"cs.AR","submitted_at":"2026-04-11T02:49:38+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"An aging-aware adaptive voltage scaling framework for AI accelerators reduces predicted threshold voltage shifts by ~19% and aging degradation by up to 46% while saving 14% lifetime power by leveraging neural network resilience.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.09073","ref_index":38,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"DRIFT: Harnessing Inherent Fault Tolerance for Efficient and Reliable Diffusion Model Inference","primary_cat":"cs.AR","submitted_at":"2026-04-10T07:56:38+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"DRIFT uses resilience analysis, targeted DVFS, and adaptive rollback ABFT to deliver 36% average energy savings or 1.7x speedup in diffusion model inference while preserving generation quality.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2603.07041","ref_index":6,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"AIReSim: A Discrete Event Simulator for Large-scale AI Cluster Reliability Modeling","primary_cat":"cs.DC","submitted_at":"2026-03-07T05:25:53+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":4.0,"formal_verification":"none","one_line_summary":"AIReSim is a discrete event simulator for evaluating failure mitigation, recovery, and capacity planning decisions in large AI clusters.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2312.11805","ref_index":20,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Gemini: A Family of Highly Capable Multimodal Models","primary_cat":"cs.CL","submitted_at":"2023-12-19T02:39:27+00:00","verdict":"CONDITIONAL","verdict_confidence":"MODERATE","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Gemini Ultra reaches human-expert performance on MMLU for the first time and sets new state-of-the-art results on 30 of 32 benchmarks, including all 20 multimodal ones tested.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null}],"limit":50,"offset":0}