https://chiosa.app/papers/accordion-elastic-depth-inference-via-fixed-prefix-layer-looping-2610-32492Chiosaen2026-10-26T20:35:07ZOne Transformer serves five compute budgets with shared weightshttps://chiosa.app/papers/accordion-elastic-depth-inference-via-fixed-prefix-layer-looping-2610-32492?lang=itChiosait2026-10-26T20:35:07ZUn Transformer offre cinque budget di calcolo con pesi condivisihttps://chiosa.app/papers/accordion-elastic-depth-inference-via-fixed-prefix-layer-looping-2610-32492?lang=frChiosafr2026-10-26T20:35:07ZUn Transformer propose cinq budgets de calcul avec des poids partagéshttps://chiosa.app/papers/agentic-ai-defense-decode-time-provenance-binding-for-mcp-2610-31974Chiosaen2026-10-26T20:35:07ZA decode-time defense keeps hidden instructions from critical callshttps://chiosa.app/papers/agentic-ai-defense-decode-time-provenance-binding-for-mcp-2610-31974?lang=frChiosafr2026-10-26T20:35:07ZUne défense au décodage bloque les instructions cachées avant les appels critiqueshttps://chiosa.app/papers/biloco-binary-low-rank-corrections-for-post-training-quantization-2610-32243Chiosaen2026-10-26T20:35:07ZBinary corrections match or improve quality at equal storagehttps://chiosa.app/papers/biloco-binary-low-rank-corrections-for-post-training-quantization-2610-32243?lang=itChiosait2026-10-26T20:35:07ZCorrezioni binarie: qualità pari o migliore a parità di spaziohttps://chiosa.app/papers/biloco-binary-low-rank-corrections-for-post-training-quantization-2610-32243?lang=esChiosaes2026-10-26T20:35:07ZLas correcciones binarias igualan o mejoran la calidad con el mismo almacenamientohttps://chiosa.app/papers/data-quality-in-rvl-cdip-a-systematic-audit-2610-32126Chiosaen2026-10-26T20:35:07ZRVL-CDIP scores reflect more than document recognitionhttps://chiosa.app/papers/data-quality-in-rvl-cdip-a-systematic-audit-2610-32126?lang=itChiosait2026-10-26T20:35:07ZI punteggi di RVL-CDIP misurano più del riconoscimento dei documentihttps://chiosa.app/papers/data-quality-in-rvl-cdip-a-systematic-audit-2610-32126?lang=esChiosaes2026-10-26T20:35:07ZLas puntuaciones de RVL-CDIP reflejan algo más que reconocer documentoshttps://chiosa.app/papers/detecting-insecure-compositions-of-clean-agent-skills-via-capability-flow-graph-analysis-2610-32155Chiosaen2026-10-26T20:35:07ZClean agent skills can combine into unsafe action chainshttps://chiosa.app/papers/detecting-insecure-compositions-of-clean-agent-skills-via-capability-flow-graph-analysis-2610-32155?lang=itChiosait2026-10-26T20:35:07ZSkill pulite per agenti possono formare catene di azioni insicurehttps://chiosa.app/papers/embedding-large-pqc-kems-in-ikev2-2610-32356Chiosaen2026-10-26T20:35:07ZQUIC speeds two IKEv2 key types, but huge keys still strugglehttps://chiosa.app/papers/embedding-large-pqc-kems-in-ikev2-2610-32356?lang=itChiosait2026-10-26T20:35:07ZQUIC accelera due tipi di chiave IKEv2, ma le chiavi enormi restano difficilihttps://chiosa.app/papers/embedding-large-pqc-kems-in-ikev2-2610-32356?lang=esChiosaes2026-10-26T20:35:07ZQUIC acelera dos tipos de clave en IKEv2, pero las claves enormes siguen dando problemashttps://chiosa.app/papers/forecasting-associations-not-interventions-a-preregistered-causal-audit-of-covariate-aware-time-series-founda-2610-31869Chiosaen2026-10-26T20:35:07ZWhat-if forecasts can mistake association for causal effecthttps://chiosa.app/papers/full-weight-vla-fine-tuning-without-an-fp32-master-copy-silently-erases-most-updates-2610-31859Chiosaen2026-10-26T20:35:07ZFull-precision weight copies help two robot models learn taskshttps://chiosa.app/papers/hyscale-scaling-long-context-training-for-learned-sparse-attention-2610-32057Chiosaen2026-10-26T20:35:07ZHyScale speeds up million-token training for sparse attentionhttps://chiosa.app/papers/hyscale-scaling-long-context-training-for-learned-sparse-attention-2610-32057?lang=itChiosait2026-10-26T20:35:07ZHyScale accelera l’addestramento con un milione di token per l’attenzione sparsahttps://chiosa.app/papers/hyscale-scaling-long-context-training-for-learned-sparse-attention-2610-32057?lang=esChiosaes2026-10-26T20:35:07ZHyScale acelera el entrenamiento con atención dispersa con un millón de tokenshttps://chiosa.app/papers/incomparable-minima-in-agent-tool-authorization-2610-31914Chiosaen2026-10-26T20:35:07ZPermission tables can offer several minimal grantshttps://chiosa.app/papers/knowing-when-to-stop-the-evidence-action-gap-in-coding-agents-2610-31895Chiosaen2026-10-26T20:35:07ZAn escalation option helps coding agents stop at repair barriershttps://chiosa.app/papers/kvstream-adaptive-parallel-storage-management-for-kv-caching-in-large-scale-llm-serving-2610-31827Chiosaen2026-10-26T20:35:07ZKVStream uses idle storage to speed long-context model servinghttps://chiosa.app/papers/loca-enabling-multi-modality-in-geospatial-location-encoders-2610-32894Chiosaen2026-10-26T20:35:07ZLoCA turns satellite imagery into location embeddingshttps://chiosa.app/papers/loca-enabling-multi-modality-in-geospatial-location-encoders-2610-32894?lang=itChiosait2026-10-26T20:35:07ZLoCA trasforma le immagini satellitari in embedding di localitàhttps://chiosa.app/papers/loca-enabling-multi-modality-in-geospatial-location-encoders-2610-32894?lang=frChiosafr2026-10-26T20:35:07ZLoCA transforme les images satellites en plongements de lieuxhttps://chiosa.app/papers/metadb-putting-file-system-structure-back-into-distributed-databases-2610-32247Chiosaen2026-10-26T20:35:07ZMetaDB reshapes databases around file-system metadatahttps://chiosa.app/papers/metadb-putting-file-system-structure-back-into-distributed-databases-2610-32247?lang=itChiosait2026-10-26T20:35:07ZMetaDB ripensa i database intorno ai metadati dei file systemhttps://chiosa.app/papers/metadb-putting-file-system-structure-back-into-distributed-databases-2610-32247?lang=esChiosaes2026-10-26T20:35:07ZMetaDB adapta las bases de datos a los metadatos de sistemas de archivoshttps://chiosa.app/papers/metadb-putting-file-system-structure-back-into-distributed-databases-2610-32247?lang=frChiosafr2026-10-26T20:35:07ZMetaDB repense les bases autour des métadonnées de fichiershttps://chiosa.app/papers/omni-woodpecker-hallucination-correction-via-evidence-prior-rebalancing-for-omni-llms-2610-27103Chiosaen2026-10-26T20:35:07ZA frozen multimodal model can rebalance conflicting cueshttps://chiosa.app/papers/on-task-off-track-benchmarking-llm-and-agent-robustness-to-near-target-prompt-injection-2610-32531Chiosaen2026-10-26T20:35:07ZRelated prompt attacks reveal missed failureshttps://chiosa.app/papers/on-task-off-track-benchmarking-llm-and-agent-robustness-to-near-target-prompt-injection-2610-32531?lang=esChiosaes2026-10-26T20:35:07ZAtaques relacionados revelan fallos que pasan inadvertidoshttps://chiosa.app/papers/on-task-off-track-benchmarking-llm-and-agent-robustness-to-near-target-prompt-injection-2610-32531?lang=frChiosafr2026-10-26T20:35:07ZDes attaques liées à la tâche révèlent des échecs manquéshttps://chiosa.app/papers/pairing-matters-performance-failures-and-efficiency-in-lightweight-coding-agents-2610-32131Chiosaen2026-10-26T20:35:07ZCoding-agent harness choice changes lightweight models’ resultshttps://chiosa.app/papers/pairing-matters-performance-failures-and-efficiency-in-lightweight-coding-agents-2610-32131?lang=esChiosaes2026-10-26T20:35:07ZEl harness cambia los resultados de los modelos ligeros de programaciónhttps://chiosa.app/papers/pairing-matters-performance-failures-and-efficiency-in-lightweight-coding-agents-2610-32131?lang=frChiosafr2026-10-26T20:35:07ZLe choix de l’agent modifie les résultats des modèles légers de programmationhttps://chiosa.app/papers/probing-the-limits-of-forward-screening-and-inverse-design-of-crystal-materials-2610-32678Chiosaen2026-10-26T20:35:07ZA unified crystal database changes screening and novelty claimshttps://chiosa.app/papers/probing-the-limits-of-forward-screening-and-inverse-design-of-crystal-materials-2610-32678?lang=itChiosait2026-10-26T20:35:07ZUn database cristallino integrato cambia screening e novitàhttps://chiosa.app/papers/probing-the-limits-of-forward-screening-and-inverse-design-of-crystal-materials-2610-32678?lang=esChiosaes2026-10-26T20:35:07ZUna base unificada de cristales cambia el análisis del cribado y la novedadhttps://chiosa.app/papers/probing-the-limits-of-forward-screening-and-inverse-design-of-crystal-materials-2610-32678?lang=frChiosafr2026-10-26T20:35:07ZUne base unifiée de cristaux revoit le criblage et les critères de nouveautéhttps://chiosa.app/papers/pseudo-labels-as-rewards-not-targets-group-relative-policy-optimization-for-semi-supervised-asr-2610-32572Chiosaen2026-10-26T20:35:07ZGroup scoring helps speech models avoid teacher errorshttps://chiosa.app/papers/pseudo-labels-as-rewards-not-targets-group-relative-policy-optimization-for-semi-supervised-asr-2610-32572?lang=itChiosait2026-10-26T20:35:07ZIl punteggio di gruppo aiuta i modelli vocali a evitare gli errori del teacherhttps://chiosa.app/papers/pseudo-labels-as-rewards-not-targets-group-relative-policy-optimization-for-semi-supervised-asr-2610-32572?lang=esChiosaes2026-10-26T20:35:07ZLa puntuación grupal ayuda a los modelos de voz a evitar errores del docentehttps://chiosa.app/papers/pseudo-labels-as-rewards-not-targets-group-relative-policy-optimization-for-semi-supervised-asr-2610-32572?lang=frChiosafr2026-10-26T20:35:07ZLe score par groupe aide les modèles vocaux à éviter les erreurs de leur enseignanthttps://chiosa.app/papers/rewriting-scaling-laws-for-synthetic-data-with-compute-and-data-constraints-2610-31867Chiosaen2026-10-26T20:35:07ZA formula predicts how much rewritten text to train onhttps://chiosa.app/papers/rynnunify-towards-unified-embodied-brain-2610-32965Chiosaen2026-10-26T20:35:07ZRynnUnify links video prediction with long-horizon planninghttps://chiosa.app/papers/rynnunify-towards-unified-embodied-brain-2610-32965?lang=itChiosait2026-10-26T20:35:07ZRynnUnify collega la previsione video alla pianificazione estesahttps://chiosa.app/papers/rynnunify-towards-unified-embodied-brain-2610-32965?lang=esChiosaes2026-10-26T20:35:07ZRynnUnify vincula la predicción de video con la planificación a largo plazohttps://chiosa.app/papers/rynnunify-towards-unified-embodied-brain-2610-32965?lang=frChiosafr2026-10-26T20:35:07ZRynnUnify relie prédiction vidéo et planification à long termehttps://chiosa.app/papers/sure-tagger-structured-speech-representations-for-compositional-evaluation-and-targeted-data-selection-2610-32198Chiosaen2026-10-26T20:35:07ZSURE-Tagger makes speech conditions easier to comparehttps://chiosa.app/papers/sure-tagger-structured-speech-representations-for-compositional-evaluation-and-targeted-data-selection-2610-32198?lang=itChiosait2026-10-26T20:35:07ZSURE-Tagger facilita il confronto delle condizioni vocalihttps://chiosa.app/papers/sure-tagger-structured-speech-representations-for-compositional-evaluation-and-targeted-data-selection-2610-32198?lang=esChiosaes2026-10-26T20:35:07ZSURE-Tagger facilita comparar condiciones del hablahttps://chiosa.app/papers/sure-tagger-structured-speech-representations-for-compositional-evaluation-and-targeted-data-selection-2610-32198?lang=frChiosafr2026-10-26T20:35:07ZSURE-Tagger facilite la comparaison des conditions vocaleshttps://chiosa.app/papers/tgr-evodag-a-dag-based-multi-agent-code-evolution-framework-for-generative-recommendation-on-tencent-2610-32326Chiosaen2026-10-26T20:35:07ZA search framework adapts recommendation pipelines across stageshttps://chiosa.app/papers/the-3ds-oracle-bin-attack-via-the-network-2610-32206Chiosaen2026-10-26T20:35:07Z3-D Secure lookup results can expose account signalshttps://chiosa.app/papers/the-3ds-oracle-bin-attack-via-the-network-2610-32206?lang=itChiosait2026-10-26T20:35:07ZI risultati del lookup 3-D Secure possono rivelare segnali sugli accounthttps://chiosa.app/papers/the-3ds-oracle-bin-attack-via-the-network-2610-32206?lang=esChiosaes2026-10-26T20:35:07ZLos resultados de consulta de 3-D Secure pueden revelar señales de cuentashttps://chiosa.app/papers/the-3ds-oracle-bin-attack-via-the-network-2610-32206?lang=frChiosafr2026-10-26T20:35:07ZLes résultats de requête 3-D Secure peuvent révéler des indices sur les compteshttps://chiosa.app/papers/the-llens-platform-for-on-premises-large-language-model-agents-in-a-japanese-university-hospital-development-2610-32151Chiosaen2026-10-26T20:35:07ZA hospital AI platform runs with its tools inside a closed networkhttps://chiosa.app/papers/the-llens-platform-for-on-premises-large-language-model-agents-in-a-japanese-university-hospital-development-2610-32151?lang=itChiosait2026-10-26T20:35:07ZUna piattaforma AI ospedaliera opera con i suoi strumenti in una rete chiusahttps://chiosa.app/papers/the-llens-platform-for-on-premises-large-language-model-agents-in-a-japanese-university-hospital-development-2610-32151?lang=esChiosaes2026-10-26T20:35:07ZUna plataforma de IA hospitalaria funciona con sus herramientas dentro de una red cerradahttps://chiosa.app/papers/tileshare-training-free-acceleration-of-overlapping-tiled-vit-decoding-2610-32233Chiosaen2026-10-26T20:35:07ZTileShare reduces repeated work in tiled vision decodershttps://chiosa.app/papers/tileshare-training-free-acceleration-of-overlapping-tiled-vit-decoding-2610-32233?lang=esChiosaes2026-10-26T20:35:07ZTileShare reduce cálculos repetidos en decodificadores visuales por mosaicoshttps://chiosa.app/papers/trusting-the-tool-asymmetric-deference-of-llm-vulnerability-auditors-to-static-analysis-evidence-2610-32876Chiosaen2026-10-26T20:35:07ZFalse safety claims can mislead code-auditing modelshttps://chiosa.app/papers/trusting-the-tool-asymmetric-deference-of-llm-vulnerability-auditors-to-static-analysis-evidence-2610-32876?lang=itChiosait2026-10-26T20:35:07ZLe false rassicurazioni sulla sicurezza possono fuorviare i modelli che verificano il codicehttps://chiosa.app/papers/trusting-the-tool-asymmetric-deference-of-llm-vulnerability-auditors-to-static-analysis-evidence-2610-32876?lang=esChiosaes2026-10-26T20:35:07ZLas falsas garantías de seguridad pueden confundir a los modelos que auditan códigohttps://chiosa.app/papers/when-should-i-trust-this-designing-and-evaluating-factuality-indicators-for-llm-based-chatbots-2610-32883Chiosaen2026-10-26T20:35:07ZHow accuracy cues shape judgments of chatbot answershttps://chiosa.app/papers/when-should-i-trust-this-designing-and-evaluating-factuality-indicators-for-llm-based-chatbots-2610-32883?lang=esChiosaes2026-10-26T20:35:07ZCómo influyen las señales de precisión en las valoraciones de chatbotshttps://chiosa.app/papers/why-full-duplex-speech-models-collapse-in-long-conversations-2610-32349Chiosaen2026-10-26T20:35:07ZTwo inference changes reduce long-conversation failureshttps://chiosa.app/papers/why-full-duplex-speech-models-collapse-in-long-conversations-2610-32349?lang=esChiosaes2026-10-26T20:35:07ZDos cambios durante la generación reducen los fallos en conversaciones largashttps://chiosa.app/papers/why-full-duplex-speech-models-collapse-in-long-conversations-2610-32349?lang=frChiosafr2026-10-26T20:35:07ZDeux changements à la génération réduisent les échecs en conversation longue