https://chiosa.app/papers/realcli-toward-a-unified-evaluation-of-human-agent-collaboration-for-cli-agents-2610-22507Chiosaen2026-10-29T19:16:34ZCLI agents can finish tasks but miss collaboration checkshttps://chiosa.app/papers/dtca-denoising-token-credit-assignment-for-reinforcement-learning-in-diffusion-language-models-2610-35201Chiosaen2026-10-29T14:37:00ZToken-level credit improves diffusion model math traininghttps://chiosa.app/papers/efficient-streaming-audio-visual-target-speaker-extraction-for-real-world-acoustic-scenes-2610-32749Chiosaen2026-10-29T14:37:00ZA broader speech test and a single-pass extractorhttps://chiosa.app/papers/omnishowcase-omnidirectional-subject-consistent-video-generation-from-multi-view-images-2610-35684Chiosaen2026-10-29T14:37:00ZA video method uses 3D shape to keep subjects consistenthttps://chiosa.app/papers/sensorimotor-alignment-a-platonic-proxy-for-embodied-policy-success-2610-31670Chiosaen2026-10-29T14:37:00ZA motion-alignment score tracks robot models’ task successhttps://chiosa.app/papers/tao-tool-aware-joint-scheduling-for-llm-based-agentic-serving-2610-37230Chiosaen2026-10-29T14:37:00ZA scheduler uses tool timing to place waiting AI sessionshttps://chiosa.app/papers/ask-your-advisor-learning-when-to-consult-stronger-models-in-agentic-systems-2610-13870Chiosaen2026-10-29T11:00:41ZARIA learns when a task agent should ask for stronger advicehttps://chiosa.app/papers/bellowskv-one-latent-for-many-layers-expanded-on-demand-2610-35023Chiosaen2026-10-29T11:00:41ZOne shared cache representation serves several model layershttps://chiosa.app/papers/card-contact-anchored-retargeting-for-dexterous-manipulation-2610-36539Chiosaen2026-10-29T11:00:41ZContact targets help robot hands reproduce human touchhttps://chiosa.app/papers/copying-is-not-grounding-training-faithful-long-form-rag-with-authored-sentence-credit-2610-18442Chiosaen2026-10-29T11:00:41ZA sentence-level reward helps retrieval models ground their own writinghttps://chiosa.app/papers/decision-models-as-judges-for-hallucination-detection-2610-37229Chiosaen2026-10-29T11:00:41ZDecision models catch some hallucinations, but confidence can misleadhttps://chiosa.app/papers/doc2harness-benchmarking-llms-on-long-horizon-harness-engineering-of-coding-agents-2610-20344Chiosaen2026-10-29T11:00:41ZTests reveal gaps in models that build coding-agent softwarehttps://chiosa.app/papers/execdiffuser-closing-the-sample-to-execution-gap-in-diffusion-planning-2610-36837Chiosaen2026-10-29T11:00:41ZBetter driving options do not always lead to better choiceshttps://chiosa.app/papers/from-detection-to-decision-calibration-and-deferral-for-prompt-injection-detection-2610-20907Chiosaen2026-10-29T11:00:41ZAttack scores alone do not settle prompt-injection decisionshttps://chiosa.app/papers/generative-representation-learning-for-scalable-latent-tokenization-2610-20533Chiosaen2026-10-29T11:00:41ZA new image tokenizer scales while resisting shortcut featureshttps://chiosa.app/papers/periodic-weak-spots-phase-sensitivity-from-chunked-kv-cache-compression-2609-36322Chiosaen2026-10-29T11:00:41ZToken position can weaken retrieval from compressed memoryhttps://chiosa.app/papers/programgym-scaling-training-environments-for-long-horizon-program-reproduction-agents-2610-13395Chiosaen2026-10-29T11:00:41ZProgramGym trains agents to rebuild programs from behaviorhttps://chiosa.app/papers/right-click-wrong-box-grading-gui-grounding-by-what-a-click-does-2610-36831Chiosaen2026-10-29T11:00:41ZBox scores miss clicks that still achieve the requested actionhttps://chiosa.app/papers/rsigen-reciprocal-self-improvement-for-agentic-visual-generation-2610-18688Chiosaen2026-10-29T11:00:41ZRSIgen links learning between image planners and generatorshttps://chiosa.app/papers/simvta-simulation-grounded-video-trajectory-alignment-for-driving-world-action-models-2610-13781Chiosaen2026-10-29T11:00:41ZRendered futures give driving models feedback on new pathshttps://chiosa.app/papers/tapestry-a-single-backbone-for-video-generation-and-robot-control-2610-36823Chiosaen2026-10-29T11:00:41ZA video model can guide robot actions without generating videohttps://chiosa.app/papers/tulva-practical-asynchronous-threshold-fhe-2610-20521Chiosaen2026-10-29T11:00:41ZTulva lets enough committee members decrypt without being online togetherhttps://chiosa.app/papers/unit-level-aggregation-breaks-coverage-in-causal-foundation-models-2610-22770Chiosaen2026-10-29T11:00:41ZJoint models improve intervals for group treatment effectshttps://chiosa.app/papers/what-makes-world-action-models-generalize-an-empirical-study-of-test-time-future-modeling-2609-34981Chiosaen2026-10-29T11:00:41ZNoisy future video can guide robot actions in one passhttps://chiosa.app/papers/when-pointing-becomes-profiling-social-bias-and-refusal-in-text-steerable-visual-encoders-2610-35017Chiosaen2026-10-29T11:00:41ZA text-guided localizer can route attention by stereotypeshttps://chiosa.app/papers/when-to-ask-for-help-a-decision-theoretic-account-and-measurement-standard-for-slow-module-scheduling-in-dual-2610-21160Chiosaen2026-10-29T11:00:41ZA value rule explains when slow corrections are worth their costhttps://chiosa.app/papers/where-does-the-performance-go-understanding-and-recovering-performance-regressions-in-c-to-rust-translation-2610-23044Chiosaen2026-10-29T11:00:41ZPerfR turns recurring C-to-Rust slowdowns into reusable ruleshttps://chiosa.app/papers/habit-bench-evaluating-habit-induction-from-longitudinal-weak-evidence-in-agent-memory-2610-24520Chiosaen2026-10-28T19:27:59ZMemory systems struggle to turn user history into sound choiceshttps://chiosa.app/papers/hy-nomas-towards-free-form-interactive-world-generation-2610-36729Chiosaen2026-10-28T19:27:59ZA video model carries scenes and actions across turnshttps://chiosa.app/papers/memory-tax-dissecting-the-net-value-of-agent-memory-2610-34073Chiosaen2026-10-28T19:27:59ZMany agent memory systems lose to direct history searchhttps://chiosa.app/papers/representation-redistribution-in-sparse-autoencoders-2610-21240Chiosaen2026-10-28T19:27:59ZA method finds feature changes that recur across promptshttps://chiosa.app/papers/spatial-audio-generation-and-editing-with-executable-plans-and-self-distilling-transfusion-2610-34415Chiosaen2026-10-28T19:27:59ZOne framework plans and edits spatial audio sceneshttps://chiosa.app/papers/work-agents-systems-capabilities-and-optimization-2610-35845Chiosaen2026-10-28T19:27:59ZA framework connects digital work agents to the work they must finishhttps://chiosa.app/papers/worldcast-world-consistent-character-video-generation-with-large-angle-camera-control-2610-33135Chiosaen2026-10-28T19:27:59ZWorldCast keeps characters coherent through large camera moveshttps://chiosa.app/papers/balanced-state-subspaces-for-mamba-2-state-compression-2610-36617Chiosaen2026-10-28T15:18:30ZMamba-2 compression preserves jointly useful state directionshttps://chiosa.app/papers/benchmeup-building-software-engineering-benchmarks-from-natural-language-specifications-2610-37283Chiosaen2026-10-28T15:18:30ZBenchMeUp turns research specifications into runnable benchmarkshttps://chiosa.app/papers/can-discrete-tokens-serve-as-semantically-lossless-representations-of-visual-signals-2610-37662Chiosaen2026-10-28T15:18:30ZA visual tokenizer keeps discrete features informativehttps://chiosa.app/papers/can-test-time-memory-evolution-outperform-test-time-training-for-scientific-discovery-2610-32046Chiosaen2026-10-28T15:18:30ZExternal memory helps models find better mathematical solutionshttps://chiosa.app/papers/channel-selection-for-mixture-of-experts-decoding-needs-no-calibration-data-2610-37386Chiosaen2026-10-28T15:18:30ZFrozen MoE decoding can select channels without calibrationhttps://chiosa.app/papers/chronoreplay-one-answer-interface-is-not-enough-for-streaming-spatial-intelligence-2610-34283Chiosaen2026-10-28T15:18:30ZChronoReplay trains video models to answer spatial questionshttps://chiosa.app/papers/closed-loop-aware-planning-via-imitation-free-speed-scoring-for-end-to-end-autonomous-driving-2610-36079Chiosaen2026-10-28T15:18:30ZA driving model learns to follow requested speedshttps://chiosa.app/papers/codecap-bench-can-code-serve-as-a-good-video-caption-2610-37341Chiosaen2026-10-28T15:18:30ZExecutable code can preserve details in video captionshttps://chiosa.app/papers/concordat-typed-sessions-for-llm-agents-2610-37461Chiosaen2026-10-28T15:18:30ZA protocol gate stopped unconfirmed airline changes under pressurehttps://chiosa.app/papers/crisp-co-designing-pruning-and-scheduling-for-efficient-moe-inference-on-edge-servers-2610-37442Chiosaen2026-10-28T15:18:30ZCRISP coordinates pruning and scheduling for edge MoEhttps://chiosa.app/papers/do-we-need-an-llm-to-know-why-sources-conflict-decision-models-for-knowledge-conflict-typing-in-rag-2610-36915Chiosaen2026-10-28T15:18:30ZMost tested models match an older language model at classifying search conflictshttps://chiosa.app/papers/efficient-quantization-training-with-stochastic-rounding-via-structured-subspace-randomness-2610-36587Chiosaen2026-10-28T15:18:30ZShared rounding thresholds speed training when errors are correctedhttps://chiosa.app/papers/evolving-agent-teams-2610-36848Chiosaen2026-10-28T15:18:30ZA supervisor helps coding teams rethink GPU kernelshttps://chiosa.app/papers/featurearc-benchmarking-coding-agents-on-evolving-features-2610-37434Chiosaen2026-10-28T15:18:30ZCoding agents often lose feature correctness across stageshttps://chiosa.app/papers/follow-the-gradient-tracing-safety-degradation-back-to-training-data-2610-37481Chiosaen2026-10-28T15:18:30ZGradient filtering removes risky training exampleshttps://chiosa.app/papers/gas-pfn-graph-aware-spatiotemporal-prior-data-fitted-networks-for-zero-shot-traffic-forecasting-2610-37598Chiosaen2026-10-28T15:18:30ZA graph-aware model forecasts traffic without retraininghttps://chiosa.app/papers/governed-sovereign-rag-for-public-administration-tripartite-epistemic-routing-facet-guided-clarification-and-2610-36561Chiosaen2026-10-28T15:18:30ZA government assistant checks generated numbers against procedureshttps://chiosa.app/papers/heads-of-security-learned-attention-gates-improve-prompt-injection-robustness-2610-37657Chiosaen2026-10-28T15:18:30ZLearned attention gates reduce prompt injection in two open modelshttps://chiosa.app/papers/how-much-can-best-of-n-help-a-world-action-model-measuring-selectable-headroom-and-realizing-it-in-closed-loo-2610-37196Chiosaen2026-10-28T15:18:30ZRobot action selection gains depend on candidate designhttps://chiosa.app/papers/hqclethe-streaming-hqc-encapsulation-for-tiny-stateless-servers-2610-36808Chiosaen2026-10-28T15:18:30ZHqcLethe streams HQC keys without per-client memoryhttps://chiosa.app/papers/infradev-can-coding-agents-develop-ai-infrastructure-and-how-can-we-tell-2610-36589Chiosaen2026-10-28T15:18:30ZHidden checks expose coding agents’ incomplete infrastructure changeshttps://chiosa.app/papers/know-then-draw-a-new-learning-paradigm-for-visual-generation-2610-37626Chiosaen2026-10-28T15:18:30ZA two-stage method separates visual learning from renderinghttps://chiosa.app/papers/learning-visuomotor-quadrotor-flight-through-dynamic-narrow-gaps-via-differentiable-simulation-2610-37557Chiosaen2026-10-28T15:18:30ZQuadrotors learn to cross moving gaps in simulationhttps://chiosa.app/papers/linmix-data-mixing-as-linear-domain-attribution-2610-37390Chiosaen2026-10-28T15:18:30ZVertex sampling helps mix many language-model data domainshttps://chiosa.app/papers/lossless-compression-of-proximity-graph-indexes-for-efficient-approximate-nearest-neighbors-search-2610-37176Chiosaen2026-10-28T15:18:30ZHNSW graph compression cuts storage with little latency costhttps://chiosa.app/papers/mle-world-predicting-machine-learning-experiment-outcomes-with-a-learned-world-model-2610-37175Chiosaen2026-10-28T15:18:30ZPredicted program feedback helps agents search for solutionshttps://chiosa.app/papers/multi-agent-reinforcement-learning-of-co-adaptive-hip-exoskeleton-control-marl-exo-for-metabolic-energy-reduc-2610-37441Chiosaen2026-10-28T15:18:30ZA hip exoskeleton lowers walking energy versus a passive device in a small testhttps://chiosa.app/papers/multi-state-linear-attention-with-generalized-second-order-routing-2610-37645Chiosaen2026-10-28T15:18:30ZMerlin helps long-context models weight useful segmentshttps://chiosa.app/papers/pailitaommsearch2-0-technical-report-scaling-generative-retrieval-to-unified-end-to-end-multimodal-search-2610-36958Chiosaen2026-10-28T15:18:30ZOne model handles three kinds of shopping searchhttps://chiosa.app/papers/pdrvla-a-vision-language-action-model-with-physical-dynamics-representations-2610-37381Chiosaen2026-10-28T15:18:30ZA robot model learns scene changes without predicting videohttps://chiosa.app/papers/pipelink-contract-guided-linking-of-gpu-kernel-components-2610-37200Chiosaen2026-10-28T15:18:30ZPipeLink lets separately built GPU kernel parts work togetherhttps://chiosa.app/papers/pneuma-a-large-scale-real-world-time-series-benchmark-for-commercial-aircraft-predictive-maintenance-2610-37563Chiosaen2026-10-28T15:18:30ZA public aircraft dataset exposes limits in early fault detectionhttps://chiosa.app/papers/pushing-dexterity-and-safety-for-in-hand-object-pose-reaching-that-works-remarkably-well-2610-37652Chiosaen2026-10-28T15:18:30ZA robot hand reaches poses while dropping fewer objectshttps://chiosa.app/papers/rendering-free-event-frame-simulation-for-fast-reinforcement-learning-of-autonomous-robot-navigation-2610-37543Chiosaen2026-10-28T15:18:30ZA rendering-free simulator trains drone navigation policies in minuteshttps://chiosa.app/papers/resple2-inertial-informed-recursive-non-uniform-spline-estimation-for-multi-sensor-odometry-with-global-refin-2610-37602Chiosaen2026-10-28T15:18:30ZRESPLE2 predicts robot motion from inertial datahttps://chiosa.app/papers/roboauto-autonomous-robot-program-search-through-multi-level-physical-understanding-2610-36533Chiosaen2026-10-28T15:18:30ZRoboAuto grounds robot programs in physical scene knowledgehttps://chiosa.app/papers/rollback-experience-backtracking-in-world-models-for-policy-improvement-2610-36068Chiosaen2026-10-28T15:18:30ZRollback helps robots learn by revisiting key actionshttps://chiosa.app/papers/scaling-laws-for-tool-use-agents-via-controllable-environment-synthesis-2610-37648Chiosaen2026-10-28T15:18:30ZMore varied tool training data helps as budgets growhttps://chiosa.app/papers/secdmp-provably-secure-data-memory-dependent-prefetching-extended-version-2610-37454Chiosaen2026-10-28T15:18:30ZA hardware defense keeps prefetchers from scanning secret datahttps://chiosa.app/papers/select-ttt-query-conditioned-selective-fast-weight-updates-for-in-place-test-time-training-2610-36794Chiosaen2026-10-28T15:18:30ZSelect-TTT directs memory updates toward relevant texthttps://chiosa.app/papers/spine-compiling-subtask-trees-into-structured-context-for-long-horizon-coding-agents-2610-37224Chiosaen2026-10-28T15:18:30ZSpine helps coding agents track nested workhttps://chiosa.app/papers/teach-reality-not-ethics-a-drop-in-continued-pretraining-corpus-installs-coupled-human-biosphere-ai-welfare-d-2610-37484Chiosaen2026-10-28T15:18:30ZTraining helps models resist pressure to approve ecological harmhttps://chiosa.app/papers/tessera-structure-completed-evidence-units-for-web-retrieval-augmented-generation-2610-37080Chiosaen2026-10-28T15:18:30ZTessera gets webpage evidence into the reader’s contexthttps://chiosa.app/papers/the-structure-gap-when-more-data-cannot-replace-structured-reasoning-2610-37523Chiosaen2026-10-28T15:18:30ZKnown task structure can resolve gaps left by exampleshttps://chiosa.app/papers/the-trinity-capability-driven-recursive-self-improvement-for-multimodal-agents-2610-36077Chiosaen2026-10-28T15:18:30ZAgents can improve their tools by grouping shared abilitieshttps://chiosa.app/papers/thinkworld-world-imagination-reinforces-chain-of-dynamics-reasoning-2610-37666Chiosaen2026-10-28T15:18:30ZRobot learning uses video to ground motion reasoninghttps://chiosa.app/papers/tops-reshaping-flow-source-and-transport-for-multi-task-robot-manipulation-2610-36694Chiosaen2026-10-28T15:18:30ZTOPS improves robot task success with task-specific flow pathshttps://chiosa.app/papers/trident-seq-profile-matched-prefix-future-contrast-for-label-efficient-payment-risk-modeling-2610-37365Chiosaen2026-10-28T15:18:30ZMatching payment histories helps risk models learn from fewer labelshttps://chiosa.app/papers/vibevideobench-evaluating-multimodal-agents-on-code-driven-video-generation-2610-35864Chiosaen2026-10-28T15:18:30ZA new test shows where video-making agents still fall shorthttps://chiosa.app/papers/vision-language-action-or-world-action-just-action-models-2610-37219Chiosaen2026-10-28T15:18:30ZRobot policies can learn from the future without generating images (F1, F3)https://chiosa.app/papers/watermark-basins-black-box-detection-of-distilled-llms-2610-37266Chiosaen2026-10-28T15:18:30ZA text-only test traces watermarked training into student modelshttps://chiosa.app/papers/when-does-perturbation-consistency-add-information-beyond-softmax-confidence-theory-and-evidence-for-the-deci-2610-36640Chiosaen2026-10-28T15:18:30ZPerturbation consistency adds little to clean image error checkshttps://chiosa.app/papers/whether-is-not-when-separating-occurrence-from-timing-in-polymarket-prices-2610-36639Chiosaen2026-10-28T15:18:30ZPrediction prices reveal more about whether than whenhttps://chiosa.app/papers/workfactory-turning-professional-intents-into-executable-work-for-scalable-agent-training-2610-37161Chiosaen2026-10-28T15:18:30ZA task-building framework checks whether work tasks can be solvedhttps://chiosa.app/papers/world-dreamer-v5-lite-a-multi-robot-world-action-model-with-one-dreamer-2610-37518Chiosaen2026-10-28T15:18:30ZA shared spatial map helps robots coordinate across viewshttps://chiosa.app/papers/your-policy-is-secretly-a-harness-the-decision-geometry-of-agent-interfaces-2610-37255Chiosaen2026-10-28T15:18:30ZA harness can hurt a frozen agent and help it after retraining