NeurIPS2026 Notes TODO

Total: 233 | Completed: 233 | Pending: 0

---

• a unified uncertainty representation for graph neural networks via doubly-spectr | arXiv: AI Safety
• action chunking proximal policy optimization with feedback correction | arXiv: Robotics
• adam under generalized smoothness with second-moment-type stochastic gradients | arXiv: 2609.37787
• adaptive inference for functionals of m-estimands | arXiv: 2609.39274
• adaptive mass-segmented kv compression for long-context reasoning | arXiv: 2605.23200
• adast adaptive coupling for spatial-temporal forecasting | arXiv: 2609.36119
• Advancing Entropy-Level Credit Assignment in RLVR via Proximal Entropy Policy Optimization | arXiv: 2609.39402
• Aegis: Generative Gradient Masking for Privacy-Preserving Medical Federated Learning | arXiv: 2609.38339
• AgentGrad: Intervention-guided Prompt Optimization for Multi Agent Systems | arXiv: 2609.08572
• AgentHop: A Diagnostic Benchmark for Agentic Multi-Hop Scientific Question Answering | arXiv: 2609.34428
• All Roads Lead to Rome: Flow-driven Multi-Anchor Exploration for Open-Environment Active 3D Mapping | arXiv: 2609.36889
• alphapareto formulaic alpha discovery with llm-guided multi-objective reinforcem | arXiv: 2609.34188
• amortized optimal transport from sliced potentials | arXiv: 2604.15114
• anchoring adversarial trajectories to data manifolds a bilevel transfer optimiza | arXiv: 2609.38991
• asirf an agentic framework for context-dependent sensitive information redaction | arXiv: AI Safety
• at fulltilt real-time open-set 3d macromolecule detection directly from tilted 2 | arXiv: 2604.10766
• attention-discounted adaptive sampler for masked diffusion language models | arXiv: 2606.10829
• audible world models spatially aware sound generation for 3d worlds | arXiv: 2609.38444
• authority bias in language models source deference and user agreement are not in | arXiv: LLM Alignment
• autoexpert automating 3d lidar annotation from expert-crafted guidelines | arXiv: 2506.02914
• bayesian optimization with fisher information geometry gradient bounds and trust | arXiv: 2609.31107
• behavioral foundation models for quality diversity | arXiv: 2609.35615
• beyond drug discovery the nanotechnology molecular optimization nmo benchmark | arXiv: Computational Biology
• beyond empirical support structured outlier generation via sinkhorn optimal tran | arXiv: 2609.31470
• beyond missing rates rethinking incomplete multi-view clustering with protocol d | arXiv: 2606.04857
• beyond normal references discriminative few-shot anomaly detection | arXiv: 2605.23231
• beyond prediction steering vlm agents with retrospective world modeling | arXiv: 2609.39101
• beyond selection token parameterization for extreme visual token compression | arXiv: Multimodal Efficiency
• beyond spatial-domain supervision a relation constrained space for multi-modal i | arXiv: 2609.38968
• bidirectional information flow bif - a sample efficient hierarchical gaussian pr | arXiv: 2505.11294
• bimogen bidirectional motion-text generation via unified masked discrete diffusi | arXiv: 2609.35407
• binding multiple modalities via multimodal wasserstein barycenter | arXiv: Multimodal VLM
• block sparse flash attention | arXiv: 2512.07011
• breaking the uniformity trap scaling video diffusion model via splitmoe | arXiv: 2609.38140
• building transformation layers for riemannian neural networks | arXiv: 2609.35436
• can circuit alignment predict ood generalization | arXiv: 2609.31996
• can linguistic reasoning vectors enhance multimodal reasoning ability | arXiv: 2609.31140
• can pixels alone reveal image origin minimax limits and learnable interfaces for | arXiv: 2609.30997
• cass contribution-aware structured sparsity for model merging | arXiv: 2609.34184
• cellmsa context modeling for single-cell representation learning | arXiv: 2609.38908
• cheap and powerful tests for supervised subspaces per-component inference for pl | arXiv: 2609.36307
• codescaler scaling code llm training and test-time inference via reward models | arXiv: 2602.17684
• commutator memory sparse path-local reading and steering in language models | arXiv: 2609.34348
• concise and logically consistent conformal sets for neuro-symbolic concept-based | arXiv: 2605.18202
• contractbench can llm agents preserve observation contracts | arXiv: 2605.17281
• contrastive representation shaping for llm unlearning | arXiv: 2601.22028
• cost-aware best-llm identification using dueling feedback | arXiv: 2609.30360
• counterfactual rollout replay forkable environments as free process rewards for | arXiv: 2609.33875
• cte-bench counterfactual trace evaluation for stateful software simulators | arXiv: Code Intelligence
• cumulative-goodness free-riding in forward-forward networks real repairable but | arXiv: 2605.06240
• d-gap improving out-of-domain robustness via dataset-agnostic and gradient-guide | arXiv: 2511.11286
• dagent evaluate-then-grow planning for deep research agents | arXiv: Multi-Agent
• data-driven soft labeling scales dna read classification to whole-body cell-type | arXiv: 2607.04987
• deep minds and shallow probes | arXiv: 2605.11448
• deeparrhythmia segment-contextualized ecg arrhythmia classification via selectiv | arXiv: Medical Imaging
• diffpts rethinking diffusion elbo for probabilistic time series forecasting | arXiv: 2609.32363
• directuv image-conditioned uv texture generation with surface-aware positional e | arXiv: 2609.34651
• distilling what matters confidence-aware selective distillation for large langua | arXiv: Model Compression
• diversity combining for multi-path llm reasoning | arXiv: 2609.38829
• drivehierarchy a benchmark for diagnosing vlm driving capabilities from open-loo | arXiv: 2609.31814
• driving video retrieval for complex queries with structured grounding | arXiv: 2606.09109
• dynamic regret in online convex optimization with indicator switching costs | arXiv: Optimization
• efficient dataset distillation for pre-trained self-supervised models via statis | arXiv: Self-Supervised Learning
• efficient dynamic algorithms for graph neural networks with non-linear propagati | arXiv: 2609.32929
• estimating and orthogonalizing unknown pre-training gradients for continual fine | arXiv: 2609.30935
• estimation of the label-noise transition matrix with performance guarantees via | arXiv: 2609.39829
• even sharper bounds for transductive learning and its applications | arXiv: 2609.28459
• evolutionary foraging in grids intermittent search dynamics emerge in finite dep | arXiv: 2609.39239
• execution guided line-by-line code generation | arXiv: 2506.10948
• exemplar2vqa a scalable exemplar-driven visual question answering generation fra | arXiv: 2609.37655
• externalized cpdag summaries improve llm causal deduction | arXiv: 2609.31071
• factorizedhmr a hybrid framework for video human mesh recovery | arXiv: 2605.14854
• finite-sample performance of gradient descent in logistic regression with gaussi | arXiv: 2606.21683
• fluxlite inference-time proposal control for discrete diffusion models | arXiv: 2609.35947
• flyaoc evaluating agentic ontology curation of drosophila scientific knowledge b | arXiv: 2602.09163
• focus benchmarking retinal model generalization from foundation vision encoders | arXiv: 2609.33158
• fractional state space transition for long sequence modeling | arXiv: 2609.36314
• from weak data to strong policy q-targets enable provable in-context reinforceme | arXiv: 2609.30391
• g2tr generation-guided visual token reduction for separate-encoder unified multi | arXiv: 2605.12309
• gazeflow from human gaze behavior to generative egocentric gaze prediction | arXiv: Human Understanding
• geometric inductive biases for semi-supervised equalization the constellation-aw | arXiv: Signal & Communication
• geowind2plan mission-time 3d urban wind prediction for energy-efficient uav plan | arXiv: Robotics
• glare generating listening heads with appropriate reactions | arXiv: 2609.40317
• goal-conditioned supervised learning for multi-objective recommendation | arXiv: 2412.08911
• gt-harmbench benchmarking ai safety risks through the lens of game theory | arXiv: 2602.12316
• ham-world soft-hamiltonian world models with selective memory for planning | arXiv: Reinforcement Learning
• handwritten text recognition lives in the high-pixel variance subspace | arXiv: 2609.35473
• hots homophily-aware temperature scaling for graph neural network calibration | arXiv: 2609.32426
• how much must a private mempool hide exact leakage thresholds for sandwich attac | arXiv: 2609.31379
• i have a stream making self-supervised learning work on continuous video | arXiv: 2609.40333v1
• i-deq a stable inertial deep equilibrium model for image restoration | arXiv: 2605.19705v2
• importance-aware obs pruning for diffusion models | arXiv: 2607.20048v3
• information bottleneck-guided adaptive hypergraph transformer for brain disease | arXiv: 2609.37220v1
• interpretable but fragile robustness of concept bottlenecks under geometric-sema | arXiv: 2609.38625v1
• its all training a fully synthetic single-stage recipe for llms | arXiv: 2609.37891v1
• kairos toward adaptive and parameter-efficient time series foundation models | arXiv: Time Series
• language-conditioned world modeling for visual navigation | arXiv: 2603.26741v2
• latent video prediction for world modeling an evaluation uncovering intriguing f | arXiv: 2605.15618v2
• learning chance-constrained mdps with bellman distributional certificates | arXiv: Reinforcement Learning
• learning where and what to restore for composite image restoration | arXiv: Image Restoration
• learning where it matters geometric anchoring for robust preference alignment | arXiv: LLM Alignment
• lemon-zest evolution-informed tokenization for efficient protein language modeli | arXiv: Computational Biology
• less supervision better generalization weakly supervised fake region localizatio | arXiv: AI Safety
• linguistic trajectory encoding for efficient long-horizon spatial memory in embo | arXiv: Robotics
• listening to the wise few query-key alignment unlocks latent correct answers in | arXiv: Interpretability
• llm alignment--utility asymmetry under semantic-preserving transformations | arXiv: 2609.32717
• llm judge validation under sparse overlap from inference to design | arXiv: 2609.31857v2
• logictree-rag logic tree-guided retrieval-augmented generation for long-form pat | arXiv: Information Retrieval & RAG
• loop-free inverse reinforcement learning via sequential value recovery with q-sc | arXiv: Reinforcement Learning
• m-plicits neural implicit surfaces via nested multiscale residuals | arXiv: 2609.28684
• markovian dynamics enforcer feasibility preserving correction on learned dynamic | arXiv: 2609.39888v1
• measuring collapse and correction in homogeneous-panel llm debate | arXiv: 2609.35279v1
• mechanism-aware ensemble conditioning for data-limited emulation of extreme even | arXiv: 2609.30746v1
• medkit evaluating knowledge integration and generalization in large language mod | arXiv: 2609.38543v1
• modeling quantum neural network gradient with reinforcement learning | arXiv: 2609.31066v1
• modeling whole-slide images as dynamic tumor microenvironment fields | arXiv: Medical Imaging
• momha multi-objective optimization of llm harnesses over accuracy safety and tok | arXiv: LLM Agent
• motion forcing a decoupled framework for robust video generation in motion dynam | arXiv: 2603.10408v2
• multabench benchmarking multimodal tabular learning with text and image | arXiv: 2605.10616
• multidimensional observer model and perceptual dimensions of human image quality | arXiv: 2609.38487
• multimodal llms outperform pathology foundation models in cross-domain histologi | arXiv: 2609.32876
• multivariate time series forecasting needs cross variable loss | arXiv: 2608.05742
• mutable transcripts mitigating context pollution through editable conversation s | arXiv: 2609.31354
• neural harmonic measure operator | arXiv: 2609.35752
• neural structural reasoner a brain-inspired architecture for reasoning over stru | arXiv: 2609.36620
• non-linear pricing restores tractability for a data seller | arXiv: 2609.36589
• nonparametric in-context learning under growing geometric complexity minimax opt | arXiv: 2609.31458
• not all tasks quantize equally fisher-guided quantization for visual geometry tr | arXiv: 2605.15828
• nrf-gs neural residual fields for expressive and compact gaussian splatting | arXiv: 2609.37115
• one threshold does not fit all languages language-conditional deferral for relia | arXiv: 2609.37861
• one view is enough in-the-wild monocular pretraining for novel view generation | arXiv: 2603.23488
• onecanvas 3d scene understanding via panoramic reprojection | arXiv: 2606.19253
• online learning via learned latent bayesian tracking | arXiv: 2609.31559
• open vocabulary domain unlearning | arXiv: 2609.31356
• openwhistle a large-scale longitudinal dataset and benchmark of bottlenose dolph | arXiv: 2609.34839
• otrope optimal transport-based robust off-policy evaluation for large language m | arXiv: 2609.36264
• panoptic scene program diffusion transformer | arXiv: 2609.31780
• parameter symmetries determine representational geometry in overparameterized no | arXiv: 2609.39078
• passing an endless journey through reconstructed spacetime with ai-generated sou | arXiv: 2609.27489
• personamanifold revealing and exploiting curved geometry in llm persona represen | arXiv: 2609.34571
• phaedra learning high-fidelity discrete tokenization for the physical science | arXiv: Physics & Scientific Computing
• phoebi an open-world benchmark for bacterial identification in phase-contrast mi | arXiv: 2606.22890
• phyprobe rethinking physical consistency evaluation in generated videos | arXiv: 2609.38377
• pisco precise video instance insertion with sparse control | arXiv: 2602.08277
• pixeldit2 representation-grounded pixel diffusion transformers | arXiv: 2609.24919
• pocketve stable and property-guided structure-based drug design with variance-ex | arXiv: Computational Biology
• polytopobench a benchmark for complex vector polygon generation from remote sens | arXiv: 2609.32856
• posebridge bridging the skeletonization gap for zero-shot skeleton-based action | arXiv: Human Understanding
• position lets strengthen verifiability if we cant enforce reproducibility | arXiv: 2609.35854
• pr-smoother simulator-preserving non-gaussian smoothing for data assimilation | arXiv: 2609.26890
• preference-guided adaptation for open-vocabulary semantic segmentation via promp | arXiv: 2609.34528
• preserving deg rankings for gene discovery in histology-based spatial gene expre | arXiv: 2609.33928
• procompnav proactive instance navigation with comparative judgment for ambiguous | arXiv: 2605.06223
• progressive memory transformer memory-aware attention for time-series | arXiv: 2609.31351
• progressive risk estimation for accident anticipation | arXiv: 2609.32811
• provable test-time scaling for beam search in llm reasoning | arXiv: 2609.38672
• pulse identifying demonstration-utility features with sparse autoencoders | arXiv: 2609.32469
• quanvi score-based variational inference via quantum maximally mixed states | arXiv: 2609.39164
• rank-constrained adaptation for reliable real-world performance | arXiv: 2602.06924
• raw-routed mixture of adapters a causal intervention for routing collapse in tim | arXiv: 2609.39445
• reasoning-trace collapse evaluating the loss of explicit reasoning during fine-t | arXiv: Reasoning
• rebalancing reference frame dominance to improve motion in image-to-video models | arXiv: 2605.19398
• recognize -- open-set comic character re-identification | arXiv: 2609.34032
• reconstructing the vocal tract with differentiable acoustic simulation | arXiv: 2609.36737
• relation-aware graph foundation model | arXiv: 2505.12027
• remember your trace memory-guided long-horizon agentic framework for consistent | arXiv: 2605.14563
• replay-buffer engineering for noise-aware quantum circuit optimization | arXiv: Reinforcement Learning
• rethinking cross-layer information routing in diffusion transformers | arXiv: 2605.20708
• rethinking personalized generation test-time alignment via factorized ranking mo | arXiv: 2609.35695
• rethinking the state update gate for long-sequence recurrent 3d reconstruction | arXiv: 3D Vision
• revisiting diffusion fine-tuning for unsupervised domain adaptation | arXiv: 2609.33716
• revitalizing medical time series with vision-informed retrieval a vision-languag | arXiv: 2609.34652
• safe score matching diffusion policies with hamilton-jacobi reachability for onl | arXiv: 2609.33337
• sage mitigating long-horizon reasoning biases via topological guidance | arXiv: 2609.30192
• satnav a scalable benchmark for long-horizon uav vision-language navigation from | arXiv: 2609.31507
• scenescaffold active scene-state construction for unified 3d scene understanding | arXiv: 2609.33518
• scir a controllable benchmark for scientific reasoning in llms | arXiv: 2606.13020
• seed self-speculative decoding via implicit encoder-decoder | arXiv: 2609.36590
• seeing speech learning visible articulatory dynamics for speech-driven 3d facial | arXiv: 3D Vision
• seg3dparts segmentation-grounded controllable part-level 3d generation | arXiv: 3D Vision
• semmsa latent semantic-aided robust multimodal sentiment analysis with incomplet | arXiv: Multimodal VLM
• sense semantic neural speech synthesis from brain dynamics via spatial graph enc | arXiv: 2609.37601
• sheafstain sheaf-theoretic schrödinger bridge for spatially and biologically coh | arXiv: 2606.11846
• simple extensions of single-objective acquisition functions and hedge strategies | arXiv: 2609.31940
• simpleevol an agent-loop framework for llm-driven automated heuristic design wit | arXiv: 2609.37172
• smile bridging continuous optimization and discrete symbolic recovery | arXiv: 2609.04639
• social choice foundations for simulation-augmented generation | arXiv: 2609.38287
• soft geometric inductive bias for object centric dynamics | arXiv: 2512.15493
• solving every step is not enough milestone oracles reveal a composition gap in l | arXiv: 2609.32235
• spanuq span-level uncertainty quantification for large language model generation | arXiv: 2607.05721
• specdrop parameter-free category-conditioned routing for modular specialization | arXiv: 2608.04084
• spectral reversal counteracting singular value bias for graph prompting | arXiv: 2609.32143
• spectral-sphere-constrained hyper-connections | arXiv: LLM Pretraining
• spherical interpolation for backward-compatible multimodal representations | arXiv: Multimodal VLM
• stabilizing the dynamic low-rank training | arXiv: 2609.32615
• starwm self-supervised trained attention routing for robust world models | arXiv: 2609.30667
• stealth apart harm together skill cascading attacks on skill-based agent systems | arXiv: 2609.30383
• stride automated evaluation of text-to-trajectory alignment across diverse conte | arXiv: 2609.34799
• structure-guided masked autoencoders for ultra-high resolution scientific image | arXiv: 2609.30682
• structured sparse memory for recurrent reasoning | arXiv: 2609.33270
• teaching video generators to remember eliciting dynamic memory for out-of-sight | arXiv: 2605.25333
• temporal gradient inversion for private trajectory reconstruction in embodied re | arXiv: 2609.30258
• the alignment illusion in multimodal large language models | arXiv: 2609.30210
• the commit-abstain circuit why language models hallucinate instead of abstaining | arXiv: 2609.32964
• the price of locality why forward-forward underperforms backpropagation | arXiv: 2609.33240
• the shape of events edge-based inductive biases via cross-domain distillation | arXiv: 2609.30478
• thousandworlds a benchmark for climate emulation of potentially habitable exopla | arXiv: 2606.18338
• timees probabilistic and deterministic time series forecasting via evolutionary | arXiv: 2609.32384
• timetok granularity-controllable time-series generation via hierarchical tokeniz | arXiv: 2605.01418
• tokenizer-generator coupling in medical image generation | arXiv: 2608.07713
• toolsearcher optimizing tool selection at scale via reinforcement learning | arXiv: 2609.30906
• topological periodicity test toppt via confidence bound of time-delay embeddings | arXiv: 2512.06324
• towards generalizable 3d anomaly detection via relational inconsistency modeling | arXiv: 2609.35059
• towards mitigating deceptive safety alignment in large reasoning models | arXiv: 2609.36254
• towards scalable context-aware single-cell spatial transcriptomics prediction fr | arXiv: 2609.36429
• towards unified dynamic face landmark detection | arXiv: 2608.10346
• trust guided decision transformer | arXiv: 2609.31586
• tt-vidt decoupling the temporal axis for efficient motion-centric video pretrain | arXiv: 2609.33419
• two-fidelity best-action identification for stochastic minimax tree | arXiv: 2606.01708
• uncertainty quantification for computer-use agents a benchmark across vision-lan | arXiv: 2606.25760
• understanding and mitigating under-confidence in gnns from the final layer | arXiv: 2505.11335
• universal cross-prompt adversarial attacks on promptable concept segmentation | arXiv: 2609.39265
• unlearningsoup is repeated tuning necessary for large language model unlearning | arXiv: 2609.37076
• verifying neural networks with reinforcement learning | arXiv: 2609.34553
• vitex-bench benchmarking high-fidelity video scene text editing | arXiv: 2609.40356
• watermarking should be treated as a monitoring primitive | arXiv: 2605.13095
• when noise meets long-tail feature-threshold dual calibration for robust pseudo- | arXiv: 2609.33668
• when one leak pays forever context binding and the price of deterring collusion | arXiv: 2609.36667
• where root cause analysis fails a retrieval-reranking decomposition | arXiv: 2609.36686
• who says what symbolic trimodal binding mechanisms in audio-visual llms | arXiv: 2609.31193
• why deterministic prm guidance underperforms in discrete diffusion reasoning | arXiv: 2609.35472
• witeness overlap directional provenance inside open-weight model families | arXiv: 2609.31784
• xbridge entity-grounded latent bridge for heterogeneous llm communication | arXiv: 2608.11676