https://chiosa.app/papers/bagualu-2-taming-dynamic-memory-fluctuation-for-efficient-moe-model-training-over-10-000-gpus-2610-23064Chiosaen2026-10-19T10:29:55ZA training system adapts memory to rare expert-load spikeshttps://chiosa.app/papers/blitz-accelerating-the-diffusion-modules-of-alphafold3-and-boltz-2-2610-23703Chiosaen2026-10-19T10:29:55ZBlitz cuts protein structure sampling to eight stepshttps://chiosa.app/papers/blitz-accelerating-the-diffusion-modules-of-alphafold3-and-boltz-2-2610-23703?lang=itChiosait2026-10-19T10:29:55ZBlitz riduce a otto i passi di campionamento delle proteinehttps://chiosa.app/papers/causal-value-attribution-of-marketing-labels-in-llm-generated-ads-a-deconfounder-approach-2610-23025Chiosaen2026-10-19T10:29:55ZA causal method traces ad engagement to chosen labelshttps://chiosa.app/papers/disentangling-language-effects-on-tool-admission-and-post-retrieval-safety-2610-23093Chiosaen2026-10-19T10:29:55ZNon-English requests more often enter retrievalhttps://chiosa.app/papers/equivalence-testing-and-run-count-design-for-stochastic-machine-learning-experiments-2610-23700Chiosaen2026-10-19T10:29:55ZA run-count method for testing training equivalencehttps://chiosa.app/papers/equivalence-testing-and-run-count-design-for-stochastic-machine-learning-experiments-2610-23700?lang=esChiosaes2026-10-19T10:29:55ZUn método para calcular ejecuciones y probar equivalenciahttps://chiosa.app/papers/equivalence-testing-and-run-count-design-for-stochastic-machine-learning-experiments-2610-23700?lang=frChiosafr2026-10-19T10:29:55ZUne méthode pour planifier les tests d’équivalence entre entraînementshttps://chiosa.app/papers/futransformer-fp8-prefill-and-int4-decode-for-llama-style-inference-on-hopper-2610-23796Chiosaen2026-10-19T10:29:55ZSeparate GPU paths speed prompt and token processinghttps://chiosa.app/papers/futransformer-fp8-prefill-and-int4-decode-for-llama-style-inference-on-hopper-2610-23796?lang=esChiosaes2026-10-19T10:29:55ZRutas separadas en GPU aceleran prompts y tokenshttps://chiosa.app/papers/jarvis-coordination-deserves-a-brain-of-its-own-in-agentic-interactive-models-2610-23804Chiosaen2026-10-19T10:29:55ZJarvis keeps background tasks aligned with live conversationshttps://chiosa.app/papers/jarvis-coordination-deserves-a-brain-of-its-own-in-agentic-interactive-models-2610-23804?lang=itChiosait2026-10-19T10:29:55ZJarvis allinea il lavoro in background alle conversazioni attivehttps://chiosa.app/papers/lemon-equivariant-learning-over-attention-maps-for-zero-shot-misbehavior-monitoring-on-new-llms-2610-22723Chiosaen2026-10-19T10:29:55ZA graph monitor transfers hallucination detection across language modelshttps://chiosa.app/papers/look-before-you-act-bridging-active-perception-and-vision-language-action-models-via-observation-aligned-data-2610-23378Chiosaen2026-10-19T10:29:55ZTraining robot policies to act after they move the camerahttps://chiosa.app/papers/look-before-you-act-bridging-active-perception-and-vision-language-action-models-via-observation-aligned-data-2610-23378?lang=itChiosait2026-10-19T10:29:55ZAddestrare le policy robotiche a proseguire dopo aver mosso la telecamerahttps://chiosa.app/papers/many-biases-one-circuit-post-training-tunes-a-pretrained-cue-routing-mechanism-2610-22937Chiosaen2026-10-19T10:29:55ZPost-training strengthens an existing circuit for following cueshttps://chiosa.app/papers/monitor-validity-failures-llm-safety-evidence-paths-at-evolving-framework-boundaries-2610-23362Chiosaen2026-10-19T10:29:55ZCanary contracts reveal where safety evidence gets losthttps://chiosa.app/papers/not-all-test-time-compute-is-evidence-model-specific-bias-in-llm-reasoning-2610-23213Chiosaen2026-10-19T10:29:55ZDia calls other models only when reasoning answers disagreehttps://chiosa.app/papers/on-the-spot-enhancing-speculative-decoding-for-diffusion-language-models-via-commit-order-prediction-2610-22543Chiosaen2026-10-19T10:29:55ZA learned drafter predicts which masked position comes nexthttps://chiosa.app/papers/peak-memory-schedules-you-can-trust-exact-realization-certified-optimality-and-the-limits-of-recomputation-fo-2610-23328Chiosaen2026-10-19T10:29:55ZReordering generated programs makes memory savings verifiablehttps://chiosa.app/papers/peak-memory-schedules-you-can-trust-exact-realization-certified-optimality-and-the-limits-of-recomputation-fo-2610-23328?lang=frChiosafr2026-10-19T10:29:55ZRéordonner les programmes générés pour vérifier les économies mémoirehttps://chiosa.app/papers/prairiefire-igniting-high-throughput-updates-in-large-scale-approximate-nearest-neighbor-search-2610-23232Chiosaen2026-10-19T10:29:55ZPrairieFire handles frequent changes in large vector indexeshttps://chiosa.app/papers/sample-mix-or-verify-allocating-test-time-compute-under-imperfect-verifiers-2610-23255Chiosaen2026-10-19T10:29:55ZA pilot helps allocate compute when verifiers errhttps://chiosa.app/papers/scalable-long-range-machine-learning-interatomic-potentials-with-electrical-response-for-atomistic-material-m-2610-23643Chiosaen2026-10-19T10:29:55ZA local atom model adds long-range forces and electrical responsehttps://chiosa.app/papers/scalable-long-range-machine-learning-interatomic-potentials-with-electrical-response-for-atomistic-material-m-2610-23643?lang=frChiosafr2026-10-19T10:29:55ZUn modèle atomique local ajoute les forces à longue portée et la réponse électriquehttps://chiosa.app/papers/shift-reliance-alignment-ranking-vision-language-models-by-the-cues-they-trust-2610-22610Chiosaen2026-10-19T10:29:55ZA model’s cue sensitivity can change which model is besthttps://chiosa.app/papers/sparsemax-dstl-sound-robustness-with-horizon-independent-gradient-allocation-for-temporal-logic-trajectory-op-2610-23837Chiosaen2026-10-19T10:29:55ZA smoother keeps violating samples in the trajectory updatehttps://chiosa.app/papers/the-ephemeral-cost-of-switching-optimizers-across-llm-training-stages-2610-22650Chiosaen2026-10-19T10:29:55ZMatching training optimizers adds little consistent benefithttps://chiosa.app/papers/think-more-or-read-more-a-token-budget-theory-of-reasoning-with-shared-context-2610-23246Chiosaen2026-10-19T10:29:55ZA token-budget method predicts when debate groups helphttps://chiosa.app/papers/tool-recommendation-for-multi-step-agents-finding-every-tool-a-task-needs-2610-23707Chiosaen2026-10-19T10:29:55ZArcade lets agents request tools from a changing cataloghttps://chiosa.app/papers/tool-recommendation-for-multi-step-agents-finding-every-tool-a-task-needs-2610-23707?lang=itChiosait2026-10-19T10:29:55ZArcade consente agli agenti di richiedere strumenti da un catalogo in evoluzionehttps://chiosa.app/papers/tool-recommendation-for-multi-step-agents-finding-every-tool-a-task-needs-2610-23707?lang=esChiosaes2026-10-19T10:29:55ZArcade permite que los agentes soliciten herramientas de un catálogo cambiantehttps://chiosa.app/papers/tool-recommendation-for-multi-step-agents-finding-every-tool-a-task-needs-2610-23707?lang=frChiosafr2026-10-19T10:29:55ZArcade permet aux agents de demander des outils dans un catalogue évolutifhttps://chiosa.app/papers/uncertainty-aware-budget-allocation-for-efficient-llm-tree-search-2610-22858Chiosaen2026-10-19T10:29:55ZLanguage model search can spend less on predictable stepshttps://chiosa.app/papers/understanding-kv-residency-in-production-scale-agentic-llm-workloads-2610-22863Chiosaen2026-10-19T10:29:55ZReusable AI session context can strain GPU memoryhttps://chiosa.app/papers/vulnsage-adjudicating-sast-reports-by-connecting-llm-reasoning-to-program-structure-2610-23035Chiosaen2026-10-19T10:29:55ZVulnSAGE checks the code evidence behind security warningshttps://chiosa.app/papers/webcoder-scaling-general-frontend-development-with-artifact-grounded-agentic-reinforcement-learning-2610-23104Chiosaen2026-10-19T10:29:55ZWebCoder trains one agent to handle changing frontend projectshttps://chiosa.app/papers/webcoder-scaling-general-frontend-development-with-artifact-grounded-agentic-reinforcement-learning-2610-23104?lang=itChiosait2026-10-19T10:29:55ZWebCoder addestra un agente a gestire progetti frontend che cambianohttps://chiosa.app/papers/what-it-is-not-what-it-says-securing-agentic-browsing-interactions-with-web-semantics-2610-22697Chiosaen2026-10-19T10:29:55ZIntact ties browser actions to the webpage elements behind themhttps://chiosa.app/papers/who-pays-for-efficient-inference-confidence-based-small-to-large-llm-routing-across-tasks-and-languages-2610-23249Chiosaen2026-10-19T10:29:55ZSmall-model confidence helps routing on some multilingual tasks