privileged information
In machine learning, privileged information means extra data a model can see during training but never at test time. Like a student who gets hints while studying but must take the exam alone. The model learns from those hints, but deployment uses only ordinary inputs, so speed and interface stay unchanged.
Across these papers, the pattern is the same: give the training process a temporary advantage, then take it away. That advantage might be future hindsight, verified sibling rollouts, motion signals, or oracle guidance. It steers learning, stabilizes objectives, or exposes hidden correlations. The payoff is better accuracy and robustness, with no extra inference cost.
- Papers
- 85
- Released code
- 1
- First seen
- Sept 2026
- Latest
- Nov 2026
84 papers in the last two months, against 1 in the two before.
Start here
Three papers to read first, in the order they came out- 1Image forensics protocol sorts AI edits into retouching, tampering, and full synthesis
A dual-channel protocol distinguishes retouched, tampered, and synthetic images beyond binary real/fake labels.
Nov 2026 - 2Medical AI segmentation models exploit size hints leaked from benchmark captions
In medical image segmentation, the benefit of text captions comes largely from 'extent leakage', not language understanding.
Nov 2026 - 3AI agents built without training trail ordinary programs on games
Weight-space programming compiles game rules into neural weights, but these networks underperform direct Python programs and resist training.
Nov 2026
How fast it is moving
Often studied with
on-policy distillationA training method where a student model learns from a teacher using its own current policy's outputs, often improving reasoning and instruction following.34 togetherreinforcement learningA machine learning method where an agent learns to make sequences of decisions by receiving rewards or penalties for its actions.12 togethercredit assignmentDetermining which actions or steps in a sequence contributed to a final outcome, often for reinforcement learning.10 togetherWho is working on it
The papers
Most central to this idea first, not most recent.
- Top Universityeess.IV
Medical AI segmentation models exploit size hints leaked from benchmark captions
In medical image segmentation, the benefit of text captions comes largely from 'extent leakage', not language understanding.
The University of Tokyo, Indian Institute of Technology Hyderabad · Nov 2026
- Big Techcs.AI
AI agents built without training trail ordinary programs on games
Weight-space programming compiles game rules into neural weights, but these networks underperform direct Python programs and resist training.
Amazon, TU Dresden · Nov 2026
- Big Techcs.CV
Image forensics protocol sorts AI edits into retouching, tampering, and full synthesis
A dual-channel protocol distinguishes retouched, tampered, and synthetic images beyond binary real/fake labels.
Google, University of British Columbia · Nov 2026
- Chinese Techcs.CV
Code simulations teach video generators realistic physics with little data
Executable code as privileged context in on-policy self-distillation dramatically improves physical plausibility in video generation.
HKUST, Alibaba Group · Nov 2026
- Big Tech
Privileged self-practice beats self-distillation for multi-turn language-model agents
Privileged Self-Practice injects task-specific instructions into failing agent rollouts to improve multi-turn agent training.
Texas A&M University, AWS AI, Amazon · Sept 2026
- Academiccs.AI
Agent certification can fail when proposal and history correlations are erased
State-only abstraction of foundation-model agents preserves proposal coverage but can destroy certifiability unless every proposal-conditioned history fiber has a common robust-safe intervention.
Hong Kong Baptist University, Beijing Normal University · Sept 2026
- Academiccs.CV
Video reasoning models improve by auditing their own use of visual clues
A model can audit its own visual reliance and focus training on under-utilized visual tokens, improving video reasoning with far fewer resources.
University of Texas at Austin · Nov 2026
- Academiccs.AI
Omitting reasoning from training data does not harm AI agents on standard benchmarks
Thought supervision in LLM agent distillation is often unnecessary because the key information is already present in the agent's action.
North Carolina State University, LG AI Research · Nov 2026
- Top Universitycs.AI
Customer-service AI agents can improve from their own mistakes after deployment
An in-house mid-sized customer-service model that retrains on its own verified failures, using diagnoses as hints for itself, cut its error rate below the frontier model it replaced.
Meituan, University of Melbourne · Nov 2026
- Top Universitycs.LG
Language models given hints make larger models’ performance more predictable
Adding graded hints to a benchmark lifts small models off zero accuracy, making downstream accuracy from large training runs far more predictable.
Stanford University · Nov 2026
- Academiccs.AI
Flow matching speeds up speculative decoding by 39% over existing drafters
Drafting speculative tokens as a short flow over the target model's hidden states speeds generation faster than autoregressive or diffusion drafters.
Yildiz Technical University · Nov 2026
- Big Techcs.CL
Language models improve using self-generated questions and answers, without human examples
Large language models self-improve using only self-generated questions and privileged reasoning, without human-curated data.
Google, Ulm University · Oct 2026
- Chinese Techcs.LG
AI assistants watching live video get better at timing their responses
A streaming video assistant trained by distilling its own outputs with privileged hints responds at the right moment with better content.
Qwen Business Unit of Alibaba · Oct 2026
- Big Techcs.RO
Robots can reorient objects from pictures without tracking their pose
A robot hand policy using only camera images and joint angles, with goals as images instead of object poses, reorients objects and transfers from simulation to real hands.
ETH Zurich, NVIDIA · Nov 2026
- Top Universitycs.LG
Activation monitors gain only modest advantage from original model computations
Activation monitors obtain only a modest benefit from native model computations, with alternative contexts recovering most performance.
University of Oxford, University of Wisconsin–Madison · Oct 2026
- Big Techcs.LG
Agents improve by distilling hindsight from their own interactions
Retrospective Self-Distillation lets agents evolve from their own trajectories using hindsight and cross-task precedents without external supervision.
Meta AI · Oct 2026
- Chinese Techcs.AI
Self-distillation choices govern learning gains and preservation of prior abilities
Self-distillation's three axes each affect acquisition and retention differently depending on task contradiction and model identity.
Huawei Technologies Co., Ltd. · Oct 2026
- Chinese Techcs.CV
Privileged motion and gravity cues improve feedforward 3D reconstruction without extra test-time cost.
Adding training-only optical flow and gravity cues makes feed-forward 3D reconstruction more accurate and robust without extra computation at inference.
Huawei Noah's Ark Lab, University of Toronto · Oct 2026
- Chinese Techcs.LG
Self-privileged critic improves value estimation for RLVR
A self-privileged critic that uses verified same-prompt rollouts as contrastive evidence improves value estimation in RL for LLMs.
Peking University, Tencent · Sept 2026
- Research Labcs.LG
Self-teaching language models improve when the teacher gets feedback too
Teacher-Evolution Self-Distillation (TESD) enhances on-policy self-distillation by improving self-teacher supervision, outperforming existing methods.
Chinese Academy of Sciences, University of Chinese Academy of Sciences · Oct 2026
- Chinese Techcs.CV
MLLMs miss text readable by squinting; diagnosis and training method recovers it
Multimodal LLMs fail to read squint text because glyph identity is weakly encoded; training with downscaled views and transcriptions lifts ANLS from 26.50 to 88.61.
Beijing University of Posts and Telecommunications, Tencent · Oct 2026
- Big Techcs.CL
Refining user representations improves the fidelity of AI user simulation
User representations optimized via hindsight distillation enable more faithful simulation than representations from GPT-5.5.
University of Illinois Urbana-Champaign, Microsoft Research · Oct 2026
- Top Universitycs.CLcode
LLMs fail multi-step math despite solving each step correctly
LLMs often solve all intermediate math milestones yet fail to combine them into a correct final solution, making composition their largest diagnosed reasoning gap.
Rutgers University, Harvard University · Oct 2026
- Academiccs.LG
Input-blind controls mimic oracle gains for layer programs in multiple-choice evaluation
Input-blind controls produce oracle headroom as large as layer programs, showing gains do not establish computation-specific benefits.
Kent State University · Oct 2026
- Chinese Techcs.CL
Translating answer-guided reasoning cuts the cost of model distillation
CoTranslator learns to translate reverse CoT to forward CoT, improving student accuracy while using less compute.
Tsinghua University, Ant Group · Oct 2026
- Chinese Techcs.CV
Decoupling skills from expert policies improves multi-capability vision-language model training
SCOPED trains capability-specific experts and distills explicit skills into a unified VLM, outperforming mixed-capability RL.
Northeastern University, Alibaba Group · Oct 2026
- Chinese Techcs.AI
Localized Context Distillation resolves the learning dilemma in interaction experience
Localized Context Distillation improves LLM learning from interaction experience by having a teacher regenerate only critical spans of each student trajectory.
Tencent, The Chinese University of Hong Kong · Oct 2026
- Chinese Techcs.AI
Shaping teacher behavior improves self-distillation for language models
TeachBack uses bidirectional preference optimization to shape teacher trajectories for improved student reasoning in self-distillation.
Huazhong University of Science and Technology, Tsinghua University · Oct 2026
- Chinese Techcs.AI
Solved mathematical mechanisms can generate verifiable training worlds for AI agents
Generating agentic RL training environments from pre-solved mathematical mechanisms yields cheap, verifiable, stateful tasks that transfer to unseen and external benchmarks.
Georgia Institute of Technology, Alibaba Token Foundry, Alibaba Group · Sept 2026
- Chinese Techcs.LG
Criterion-aware credit assignment improves rubric-based reinforcement learning for language models
CoCA improves rubric-based reinforcement learning by 4.2% on medicine and science benchmarks through criterion-aware credit assignment.
Zhejiang University, Ant Group · Oct 2026
- Big Techcs.LG
Joint on-policy learning improves self-distillation by aligning teacher with student
A privileged teacher improves RL students only if its distillation update aligns with the student's reward gradient, motivating joint on-policy training.
Perplexity, NVIDIA · Oct 2026
- Big Techcs.AI
Mined agent skills need domain-specific evidence and format choices
Mined skill performance depends on task domain; workflows beat ontologies on one benchmark, outcome labels help on another, but no configuration wins everywhere.
Massachusetts Institute of Technology, Microsoft Corporation · Oct 2026
- Independentcs.CL
On-policy distillation can transfer think-mode advantage to no-think responses via trace-response routing
ThinkOPD outperforms uniform on-policy distillation by combining reward gain and trace-response divergence for routing.
Oct 2026
- Big Techcs.RO
Robots improve reusable manipulation skills through guided simulation practice
RPG improves robot task success from 28.6% to 95% through autonomous practice without updating model weights.
UC Berkeley, MIT · Oct 2026
- Chinese Techcs.CL
Teacher-guided training improves specialization and coordination in multi-agent models
MAS-OPD achieves the highest mean score on code and math benchmarks by using on-policy distillation with role-advantage specialization and privileged attribution.
University of Science and Technology of China, Tencent · Sept 2026
- Chinese Techcs.LG
Models can absorb spatial reasoning from verified agent traces
SpatialOPSD uses on-policy self-distillation from verified coding agent traces to give MLLMs tool-free spatial reasoning, beating SFT and GRPO.
Alibaba Group · Oct 2026
- Big Techcs.CV
Last-frame layouts guide video generation through large camera moves
LIFT enables specifying layout of future views in video generation via on-policy self-distillation from a dense-layout teacher, improving controllability under large viewpoint changes.
UC San Diego, University of Virginia · Oct 2026
- Chinese Techcs.AI
Self-distillation trains GUI agents for longer, memory-dependent tasks
GUI-SD-v2 extends on-policy self-distillation to multi-turn GUI agents, outperforming state-of-the-art methods.
Institute of Information Engineering, Chinese Academy of Sciences, Tencent · Sept 2026
- Top Universitycs.CR
AI agents turn binary patches into exploits, but only 20% succeed
AI agents can turn binary patches into working exploits for real vulnerabilities in a controlled setting.
National University of Singapore, UC Riverside · Nov 2026
- Academiccs.CV
Systematic audit reveals pervasive data quality issues in RVL-CDIP document classification benchmark
RVL-CDIP has significant label errors, test-train duplicates, and spurious cues that inflate reported classification accuracies.
Vanderbilt University · Oct 2026
- Big Techcs.RO
Humanoid robots open round doorknobs without real training
Tactile-driven motor adaptation substantially outperforms standard methods for zero-shot sim-to-real humanoid doorknob opening.
Amazon, Rice University · Nov 2026
- Big Techcs.CV
Driving simulators follow planned paths more faithfully with overhead views
BirdDreams jointly generates perspective and bird's-eye-view videos to create reactive driving simulations from a single image.
NVIDIA, KAIST · Nov 2026
- Big Techcs.CV
AI models struggle to follow events across directions in panoramic video
Panoramic video reasoning models can improve by 12.5 points through sphere-time grounded reasoning training.
Nanjing University, NVIDIA · Nov 2026
- Top Universitycs.RO
Coding agents control simulated robots less reliably under realistic constraints
The strongest coding agent succeeds only 34% of the time on robot tasks under real-world constraints.
Northwestern University, Carnegie Mellon University · Nov 2026
- Big Techcs.RO
Robots can learn door lifting with ten times fewer demonstrations
Reusing a frozen reinforcement-learned action decoder and learning a visual latent interface makes dexterous manipulation from visual demonstrations far more data-efficient.
University of California, Berkeley, Amazon · Nov 2026
- Chinese Techcs.AI
Reusing past attempts to steer data collection trains better coding agents
History-guided trajectory collection that avoids exposing prior attempts to the agent improves training data for code agents.
University of Connecticut, Ant Group · Nov 2026
- Top Universitycs.RO
Robot agents using open models can beat a published proprietary baseline
An open-weight agent, using a unified harness without privileged inputs, surpasses a proprietary model on a robot manipulation benchmark.
Tsinghua University, UC Berkeley · Nov 2026
- Chinese Techcs.CL
Language models can understand others’ beliefs yet fail to act accordingly
Language models' theory-of-mind belief inference does not guarantee optimal planning or task success in verifiable collaborative tasks.
The Chinese University of Hong Kong (Shenzhen), ByteDance Inc. · Oct 2026
- Big Techcs.AI
Language models can improve by learning from their weakened copies
Contrasting a model with its weakened copy improves reasoning accuracy without privileged information.
Massachusetts Institute of Technology, NVIDIA · Oct 2026
- Chinese Techcs.LG
Guided self-exploration lets language models teach themselves better reasoning
SEPO uses labels to guide LLM reasoning exploration, improving self-evolution efficiency and final performance.
Renmin University of China, Peking University · Oct 2026
- Chinese Techcs.LG
Time-resolved diagnostic separates irreducible velocity ambiguity from reducible training error
A time-resolved Bayes-risk decomposition separates irreducible velocity ambiguity from reducible error, guiding flow-matching time sampling that improves velocity loss and feature Fréchet distance.
Beijing Institute of Technology, ByteDance · Oct 2026
- Big Techcs.CL
Dense progress rewards help train agents on longer computer tasks
Latent-world task generation plus dense progress rewards makes long-horizon agent training tractable and improves open models across in-domain and transferred terminal benchmarks.
Adobe Research, University of California, Los Angeles · Oct 2026
- Top Universitycs.LG
Coding agents can learn neglected behaviors without sacrificing baseline performance
Combining per-turn textual feedback self-distillation with RL stabilizes training and improves hard-to-optimize agent behaviors.
Cursor Research, UC Berkeley · Nov 2026
- Chinese Techcs.CL
Outcome verification reveals failures in social agent plans that conversation ratings miss
Social agents that sound successful often produce invalid plans, while paired-world outcome training improves verified coordination.
Alibaba Group, The University of Tokyo · Oct 2026
- Big Techcs.RO
Phone scans can personalize robot pickup policies through simulation
A phone scan and object images can replace real demonstrations for personalizing a VLA policy, yielding 63.9% real-world pickup success.
Georgia Institute of Technology, Waymo LLC · Oct 2026
- Top Universitycs.CL
Gold-conditioned teacher softens targets for language model fine-tuning
Gold-Conditioned Distillation fine-tunes a model by matching a gold-conditioned teacher's soft targets, improving new-task accuracy over SFT and preserving general capabilities.
King Abdullah University of Science and Technology, Shanghai Jiao Tong University · Oct 2026
- Top Universitycs.CV
Predicting velocity fields helps video models generate more realistic fluids
FluidGen generates physically plausible fluid videos by jointly predicting a 3D velocity field alongside RGB.
Stanford University · Oct 2026
- Chinese Techcs.LG
Hindsight assigns token credit from one verified language-model rollout
HP-RLVR assigns token credit from a single rollout per prompt, training 1.21x to 2.30x faster than GRPO with comparable accuracy.
South China University of Technology, Tsinghua University · Oct 2026
- Big Techcs.CV
Self-distillation helps vision-language models reason with far fewer visual tokens
Visual token pruning degrades VLM performance primarily due to unreliable use of remaining visual evidence, not irreversible information loss.
University of Toronto, Vector Institute · Oct 2026
- Chinese Techcs.CL
Hierarchical supervision allocation improves long-horizon model distillation
Hierarchical supervision allocation via Locate-Validate-Refine improves long-horizon on-policy distillation for LLM agents.
Ant Group, Alibaba International Digital Commerce Group · Sept 2026
- Big Techcs.AI
General instructions outperform gold answers in on-policy context distillation
Using general instruction privileges instead of instance-specific gold improves out-of-distribution performance in On-Policy Context Distillation.
Columbia University, Amazon Web Services · Oct 2026
- Big Techcs.SE
New benchmark tests AI's ability to spot invalid code reviews
AI agents can judge if a code review comment is technically correct by gathering repository evidence, outperforming general LLMs.
University College London, Amazon · Sept 2026
- Academiccs.RO
Predictive visual models improve action generation for generalist robot manipulation
InternW0-Δ jointly models visual dynamics and action generation for robot manipulation, outperforming prior methods across benchmarks.
Shanghai AI Laboratory · Sept 2026
- Big Tech
Co-evolving teacher and student models improves mathematical reasoning
A recursive distillation method with a co-evolving privileged teacher improves reasoning accuracy more than using a frozen teacher.
Meta AI, University of California, Riverside · Sept 2026
- Top Universitycs.AI
Privileged supervision improves action-level credit for language-model agents
TASPO converts privileged process supervision into outcome-grounded action credit, improving agentic policy optimization by 10.6% over GRPO across three benchmarks.
Zhejiang University · Sept 2026
- Industrycs.AI
Auditing passing trajectories exposes and repairs unearned benchmark passes
Unearned passes in agentic benchmarks can be detected via trajectory auditing; patch-only repairs are insufficient because the information channel may remain open.
Scale AI · Oct 2026
- Academiccs.AI
Realistic simulated users improve training for interactive language agents
MIMESIS learns realistic user behaviors from human conversations and reasoning traces, enabling agents to train more effectively and generalize across unseen users.
Meta Superintelligence Labs, New York University · Oct 2026
- Academiccs.LG
Agent improves by distilling its own verified experiences during deployment
Deployed LLM agents can improve by self-distilling their own verified execution trajectories into persistent weight updates without a separate training phase.
University of New South Wales (UNSW Sydney) · Oct 2026
- Academiccs.LG
Anchored hypergraphs provide stable credit assignment for multi-agent teams
HySTAR uses a stable anchored hypergraph to separate credit assignment from adaptive representations, improving MARL performance across multiple benchmarks.
University of Electronic Science and Technology of China, Independent Researcher · Sept 2026
- Big Techcs.LG
Attention distillation complements token-level supervision for reasoning models
On-policy attention self-distillation (OPASD) improves reasoning accuracy by 4.98 to 8.40 percentage points while reducing training compute by 72.6% compared to token-only distillation.
ACI PLC, University of Dhaka · Sept 2026
- Chinese Techcs.CV
Real-time GUI feedback improves training for computer-use agents
ComputerSD uses real-time GUI feedback to improve computer-use agent training via online self-distillation.
Zhejiang University, Ant Group · Oct 2026
- Academiccs.LG
AI teacher adapts its hints as a language model learns, boosting performance across four domains
AutoScaffold uses a frozen teacher LLM to automatically design adaptive training scaffolds that improve language model RL performance across domains.
University of Minnesota, Twin Cities, JPMorgan AI Research · Nov 2026
- Big Techcs.RO
Driving models can recognize hazards yet fail to brake
Driving VLAs can encode and describe hazards yet ignore them when planning; inference-time hazard guidance restores braking without retraining.
Politecnico di Milano, Stanford University · Oct 2026
- Big Techcs.RO
Robots learn dexterous skills from a single egocentric human video
HandMimicX learns dexterous manipulation from a single human video using object motion and interaction priors for spatial generalization.
Tsinghua University, TsingTec · Nov 2026
- Chinese Techcs.AI
Feedback criteria that track a model's growing ability halve its training steps
Co-evolving a reward rubric alongside a policy by tracking its capability frontier improves and accelerates open-ended LLM training.
University of Science and Technology of China, Alibaba Group · Oct 2026
- Chinese Techcs.CL
Adaptive training harness keeps rubric rewards useful as policies improve
HarnessRL co-evolves task sampling, rollout guidance, and rubric criteria during RL training, beating baselines by 2.8-3.1 points on open-ended tasks.
Alibaba Token Hub, Alibaba Group, Université de Montréal · Oct 2026
- Big Techcs.RO
Adaptive sampling helps million-environment robot training solve harder tasks
Sampling task configurations where a policy has moderate success rates lets massively parallel robot RL solve harder tasks.
University of Washington, NVIDIA · Oct 2026
- Top Universitycs.LG
Divergence choice in distillation controls student model entropy
Divergence choice in distillation controls student entropy, with forward KL inflating and reverse KL deflating it.
Stanford University · Oct 2026
- Big Techcs.LG
Predictive sufficiency quantifies when context encoders aid model-based RL agents
Predictive sufficiency measures when an agent's history already identifies the environment context, making context encoders unnecessary.
King AI Labs, Microsoft Gaming, KTH Royal Institute of Technology · Oct 2026
- Big Techcs.RO
Counterfactual videos teach humanoids to handle unfamiliar objects
Counterfactual video generation turns a handful of real videos into a diverse training set that lets a humanoid pick up, carry, and drop novel objects zero-shot.
Amazon FAR, UC Berkeley · Sept 2026
- Big Techstat.ML
New hinge losses ensure optimal routing in adaptive AI classification
Cost-sensitive hinge losses enable Bayes-consistent joint training of classifiers and routers.
University of Oxford, McGill University · Oct 2026
- Big Techcs.LG
Activation steering strengthens token-level supervision for self-distilled reasoning
Activation-Conditioned Self-Distillation steers a frozen self-teacher using contrasts from verified correct trajectories, improving math and code accuracy beyond reference-conditioned self-distillation.
Rensselaer Polytechnic Institute, IBM Research · Oct 2026
- Big Techcs.CV
Wiring details determine whether affordance heads help or hurt VLA policies
Stop-gradient affordance head with active initialized residual bridge boosts VLA policy success to 96.2% on LIBERO.
Drexel University, Virginia Tech · Oct 2026
- Chinese Techcs.LG
Outcome-guided self-distillation improves LLM reasoning by adapting supervision
Outcome-guided divergence selection and prefix truncation improve on-policy self-distillation for LLM reasoning, consistently beating vanilla OPSD across model scales.
Southeast University, Huawei Noah’s Ark Lab · Oct 2026
- Academiccs.LG
A recalibration method stabilizes LLM self-distillation by treating correct and incorrect outputs differently
FIRE uses Fisher information to recalibrate supervision, stabilizing feedback-based self-distillation of LLMs.
Purdue University, Yonsei University · Oct 2026
Concepts are extracted from each paper and reused across the corpus, so this page grows on its own as the desk reads.