confidence calibration
When a model says it is sure about an answer, we want that to mean it is likely correct. This is the idea of confidence calibration: making the model's expressed certainty match its true accuracy. Poor calibration means a model can be very confident in wrong answers, which is risky in real-world use. Researchers study this to build more trustworthy AI.
The papers show that models are often overconfident, especially on hard questions or new domains. While techniques like using internal model states or test-time adjustments can improve calibration, these gains frequently do not transfer across tasks. A key difficulty is that calibration methods may teach models to predict accuracy on familiar data without fixing overconfidence on novel, challenging cases.
- Papers
- 92
- Released code
- 2
- First seen
- Sept 2026
- Latest
- Nov 2026
92 papers in the last two months, against 0 in the two before.
Start here
Three papers to read first, in the order they came out- 1AI moderators need different signals for human disagreement and costly mistakes
A cheap content moderation model matches frontier LLM accuracy but reveals two separate uncertainties.
Oct 2026 - 2Extra attempts help small language models solve terminal tasks, but not verify solutions
Small language models fall behind frontier models on terminal tasks not only in solving them but also in verifying solutions and anticipating their own success.
Oct 2026 - 3Language models can learn facts while misjudging what they know
Large language models need explicit calibration of answerability judgments after knowledge updates to avoid confident wrong answers.
Nov 2026
How fast it is moving
Often studied with
large language modelA neural network trained on massive text corpora to generate coherent text and perform diverse language tasks.16 togetherverificationChecking outputs for correctness before accepting them, often to catch errors or reward hacking.16 togethermodel routingDirecting each query or subtask to the most suitable model to optimize cost, accuracy, or quality.12 togetherWho is working on it
The papers
Most central to this idea first, not most recent.
- Chinese Techcs.AI
Language models can learn facts while misjudging what they know
Large language models need explicit calibration of answerability judgments after knowledge updates to avoid confident wrong answers.
City University of Hong Kong, Tencent Jarvis Lab · Nov 2026
- Big Techcs.CL
AI moderators need different signals for human disagreement and costly mistakes
A cheap content moderation model matches frontier LLM accuracy but reveals two separate uncertainties.
Google · Oct 2026
- Big Techcs.CL
Extra attempts help small language models solve terminal tasks, but not verify solutions
Small language models fall behind frontier models on terminal tasks not only in solving them but also in verifying solutions and anticipating their own success.
Microsoft Research, University of California, Los Angeles · Oct 2026
- Industrycs.AI
AI models best at spotting their own errors are worst at confidence
Large language models that are better at confidently telling their own correct solutions from incorrect ones tend to be worse calibrated.
AWS GenAI Innovation Center · Nov 2026
- Independentcs.LG
Language models’ own confidence becomes harder to outperform as they improve
Performance prediction only beats a model's own confidence when the model is suboptimal or the predictor adds significant capacity.
Oct 2026
- Research Labcs.CL
Post-training alignment makes language models more confidently wrong
Alignment training, not just missing knowledge, is a primary cause of LLMs' confident hallucinations, multiplying high-confidence factual errors by 10x to 35x; bounded-margin DPO cuts them.
Institute of Information Engineering, Chinese Academy of Sciences, School of Cyber Security, University of Chinese Academy of Sciences · Sept 2026
- Big Tech
Audio LLMs cannot assess their own transcription reliability.
Audio-encoder representations predict whether an Audio LLM will transcribe degraded speech reliably far better than self-assessment or tested quality and uncertainty baselines.
Apple, University of California San Diego · Sept 2026
- Independentcs.CL
Local AI can cut database agent costs without reducing accuracy
LuthorSQL answers nearly half of text-to-SQL queries locally, cutting API cost by 33.3% at the same accuracy.
Nov 2026
- Independentcs.CL
AI models answer financial facts they claim not to know
Self-reports, stated cutoffs, and date instructions all fail to prevent LLMs from leaking future knowledge into financial backtests.
Nov 2026
- Top Universitycs.LG
Controlled rewrites help LLM judge panels ignore superficial flaws
VoirDire uses controlled rewrites of inputs to reduce LLM judges' shared biases about surface features like length or formatting.
Princeton University, University of Wisconsin–Madison · Nov 2026
- Chinese Techcs.AI
Recycling high-scored unexposed candidates boosts feed discovery without engagement loss
Recycling high-scored items that the ranker predicts well but were never shown boosts discovery without harming engagement.
Alibaba · Nov 2026
- Top Universitycs.AI
Backtesting method lets AI agents forecast past events for instant feedback
FABench reveals retrieval helps agents match market forecasts, yet evidence weighing remains a critical failing.
Raven Labs, Zhejiang University · Nov 2026
- Research Labcs.LG
AI classifiers can cut robustness checking costs by weakening individual guarantees
A proxy certificate with marginal guarantees reduces robustness certification cost to a single model forward pass.
CISPA Helmholtz Center for Information Security, University of Cologne · Nov 2026
- Chinese Techcs.CL
High probe accuracy does not guarantee a usable agent alarm
High-AUROC activation probes do not establish deployable alarm policies without semantic-unit evaluation, independently calibrated thresholds, transport, and acceptable low-false-alarm utility.
Alibaba Cloud Computing · Oct 2026
- Big Techcs.LG
Training LLM forecasters across dates improves search and probability estimates
Training an LLM forecaster to search evidence and report probabilities at multiple pre-resolution dates improves forecast accuracy and calibration.
Amazon Data & Models, Georgia Institute of Technology · Oct 2026
- Research Labcs.CL
LLMs trained to predict their own accuracy learn two distinct types of confidence
LLMs trained to predict their own accuracy learn to track output consistency, not true accuracy, on unfamiliar questions.
INSERM, ENS PSL · Oct 2026
- Big Techcs.AI
Small surrogate models can flag errors in black-box LLM agents
A small surrogate model's log-probabilities audit black-box LLM agent actions better than the agent's own confidence or self-consistency.
Amazon · Oct 2026
- Independentcs.AI
Strong language models often fail to recognize their own errors
LLMs' confidence reports poorly discriminate errors from correct answers on hard problems, even with self-review.
Oct 2026
- Big Techstat.ML
CARE learns from human review while preserving calibrated risk control
CARE maintains risk control during online learning from selective human feedback while cutting human-review queries by 25-81% across four safety-critical datasets.
MIT, NVIDIA · Oct 2026
- Big Techcs.CL
Internal confidence probe eclipses verbalized scores in calibration for reasoning models
Internal hidden state probes provide well-calibrated confidence for reasoning models and can distill that calibration into the model's outputs.
NetEase, Amazon · Oct 2026
- Chinese Techcs.LG
Test-time calibration learning improves LLM accuracy and confidence without labels
An LLM's confidence can be calibrated at test time without labels using self-supervised rewards from multiple generated answers.
Hong Kong Baptist University, Huawei Noah's Ark Lab · Oct 2026
- Big Techstat.ML
Harmonic weighting makes online calibration polynomial in forecast dimension
Harmonic weighting achieves ε-calibration in d^{O(1/ε)} rounds for multi-event and multi-class forecasting.
Institute for Advanced Study, Google Research · Oct 2026
- Research Labcs.AI
Jailbreak prompts improve uncertainty calibration in black-box reasoning models
Jailbreak prompts, repurposed for uncertainty estimation, significantly improve confidence calibration in black-box large reasoning models.
Petscraft, Inria · Sept 2026
- Big Techcs.CL
Stated confidence tracks language models’ internal answer probabilities
Verbalized probabilities can be used to probe a large language model's internal probability distribution.
Apple · Oct 2026
- Chinese Techcs.IR
Direct decision scoring makes search agents faster and better calibrated
SearchJev separates bounded search judgments from generative reasoning, delivering faster, better-calibrated decisions while improving search speed and answer accuracy.
Huawei Technologies Co., Ltd., Leiden University Code Models · Oct 2026
- Chinese Techcs.LG
Differentiable confidence readouts improve reasoning accuracy and calibration together
CREDO improves reasoning accuracy and confidence calibration by replacing sampled confidence text with a deterministic, differentiable token-probability readout.
University of Science and Technology of China, Qwen Business Unit of Alibaba · Sept 2026
- Chinese Techcs.CV
Confidence scores for generated video intervals enable controllable selection without external verification
Interval-level confidence scores trained with offline verifiers and reinforcement learning improve candidate selection in generative video temporal grounding.
Alibaba Group, Beihang University · Oct 2026
- Chinese Techcs.LG
Asymmetric certainty gains from optimization hinder multimodal classification
Optimization yields asymmetric certainty gains in multimodal learning, with the strong modality more confident than the weak one, harming performance.
Nanjing University of Science and Technology, Alibaba Group · Oct 2026
- Top Universitycs.CL
Fast decision models excel locally but falter in extended systems
General decision models rival LLMs on evidence-based choices but overstate probabilities, degrade in long-horizon tasks, and can be reproduced from open-weight LLMs.
Shanghai Innovation Institute, Fudan University · Oct 2026
- Top Universitycs.AI
Decision-only judge matches reasoning judge at fraction of cost using confidence-based escalation
Confidence-gated JEV judging matches or exceeds GPT-6 accuracy on tested cascades at lower cost, but routing weakens on style-adversarial and reference-free prose.
Carnegie Mellon University · Oct 2026
- Research Labcs.CL
A modular abstention method separates accuracy training from abstention decisions
Training for accuracy then using a separate confidence estimator for abstention achieves strong and controllable abstention performance.
University of Chinese Academy of Sciences, State Key Laboratory of AI Safety, Institute of Computing Technology, CAS · Oct 2026
- Academiccs.AI
Reasoning makes language models repeat the same wrong answers
Reasoning makes independently sampled wrong answers converge, undermining self-consistency, while confidence-weighted voting fails to improve on plain majority voting.
Oklahoma State University · Sept 2026
- Big Techcs.AI
Small decision models can safely handle routine agent decisions with little loss of success
Replacing most LLM calls with a decision model cuts agent cost and latency with minimal success loss.
University of Illinois Urbana-Champaign, Princeton University · Oct 2026
- Independentcs.CL
General-purpose LLMs already work as Jev-style decision models out of the box
Pre-trained LLMs are inherently effective Jev-style decision models out of the box.
Oct 2026
- Industrycs.LGcode
Training Forecasters to Search Improves Calibration and Reduces Tool Use
Training language models to gather their own evidence during reinforcement learning improves forecasting and beats frontier models at lower cost.
Future Principle · Oct 2026
- Top Universitycs.AI
Teaching Models Confidence Can Shorten Reasoning Without Explicit Stop Training
Self-supervised confidence fine-tuning reduces generated tokens by up to 25% at matched accuracy without directly optimizing for shorter reasoning.
University of Maryland, AI Foundations, Capital One · Sept 2026
- Research Labcs.LG
Event graphs improve LLM stock forecasts across two major markets
Temporally grounded, reliability-calibrated historical event transitions add incremental stock-forecasting signal when applied as a residual correction to a multi-view LLM forecast.
Zircon Security, Utrecht University · Oct 2026
- Independentcs.LG
Random Fourier features make Gaussian process attention linear-time and well calibrated
Random Fourier feature Gaussian process attention provides linear-time complexity and calibrated uncertainty for transformer models.
Oct 2026
- Research Labcs.CL
Companion model sharpens LLM confidence without harming task performance
CoCal trains a lightweight companion on rollout hidden states to calibrate LLM confidence without sacrificing task performance.
State Key Laboratory of AI Safety, Institute of Computing Technology, Chinese Academy of Sciences · Sept 2026
- Big Techcs.AI
AI agents can retrieve relevant memories yet confuse similar names
Persistent agent memory works better when it preserves every fact at write time and resolves entity, time, and answerability questions at read time.
Amazon · Nov 2026
- Industrycs.AI
Debate traces yield useful training data when confidence replaces voting
AbCD converts multi-agent debate traces into preference data by clustering answers and ranking by confidence, improving reasoning without gold labels.
Upwork Inc. · Nov 2026
- Top Universitycs.LG
Same answer from different prompts does not mean same belief updating in language models
Matched predictions from different belief-setting routes produce different belief-update dynamics in language models.
University of Tokyo, RWTH Aachen University · Nov 2026
- Academiccs.SE
Code plagiarism detectors fail on whole repositories rewritten by AI agents
Current provenance detection methods cannot reliably trace repository-scale AI-rewritten code to its source.
Emory University, University of Illinois Urbana-Champaign · Nov 2026
- Chinese Techcs.CV
AI video generation gains emotional fidelity when rewards trust only reliable cues
EmoReward adaptively calibrates reward weights in emotional video generation using reliability learned from previous training run traces.
The Hong Kong University of Science and Technology, Tencent Hunyuan · Nov 2026
- Chinese Techcs.RO
Robot action checks trade off error detection against pinpointing causes
World models trained on clean transitions detect and localize silent action-interface mismatches that are numerically legal.
Yale University, Ant Group · Nov 2026
- Big Techcs.SE
AI agents resist mistaken criticism better when they record evidence first
Pre-feedback claim checklists prevent LLM agents from falsely reversing correct work when users complain.
Google · Oct 2026
- Chinese Techcs.CL
Language models’ self-assessments can look accurate without predicting future performance
Language models can report supplied statistics exactly yet still fail to build trustworthy domain-level self-assessments from noisy user feedback.
The Chinese University of Hong Kong (Shenzhen), Shenzhen Research Institute of Big Data · Oct 2026
- Research Labcs.AI
Debate Corrects Human Judgments, but Usually Fails Weaker AI Judges
Humans correct their judgments through debate, but weaker AI models do not.
Università degli Studi di Genova, FAIR · Oct 2026
- Research Labcs.CL
AI reviewers reward confident writing even when the science stays unchanged
LLM peer reviewers systematically reward high-confidence rhetoric regardless of scientific content, risking biased review outcomes.
CISPA Helmholtz Center for Information Security, German Research Center for Artificial Intelligence (DFKI) · Oct 2026
- Chinese Techcs.LG
Language models solve more problems when training accounts for confidence
SCARL reshapes advantages to align policy confidence with outcomes, improving reasoning accuracy and exploration in language model training.
Baidu, Nanjing University · Oct 2026
- Independentcs.CL
AI agents cut computation by a fifth while keeping accuracy competitive
A metacognitive controller reduces diffusion language agent inference FLOPs by 20.6% without losing accuracy.
Oct 2026
- Chinese Techcs.AI
Language models can give identical answers up to 2.78 times faster
Filtering low-quality branches from a candidate tree before expensive verification speeds up tree-based speculative decoding without changing outputs.
Tongji University, Nanyang Technological University · Oct 2026
- Chinese Techcs.CV
Geometry-aware test-time scaling improves camera control in video diffusion without retraining
Training-free geometry-aware candidate selection at test time improves camera control and geometric consistency in video diffusion models without retraining.
University of Freiburg, Huawei Heisenberg Research Center (Munich) · Oct 2026
- Top Universitycs.AI
Typed selective control cuts strong-model calls while preserving agent success
A typed, confidence-gated decision layer cuts strong-LLM calls while preserving agent success, but offers little advantage when cheap generative routing is already accurate.
Nanyang Technological University · Sept 2026
- Top Universitycs.RO
Predicting future object motion helps robots avoid emerging hazards
Predictive Semantic Safety uses vision-language models and conformal prediction to filter robot commands against predicted future hazards, achieving 99.3% safe episodes.
University of Michigan, King Abdullah University of Science and Technology (KAUST) · Sept 2026
- Academiccs.CR
One fake news article reverses over 74% of AI forecasts that opposed it
A single fabricated news article reverses 74-82% of forecasts that oppose it in LLM forecasting systems, and current defenses fail.
Northwestern University · Nov 2026
- Top Universitycs.AI
Cancer AI models follow false statements about their own drug resistance findings
Language models recommend cancer therapies despite genomic evidence of resistance in up to 81.4% of cases.
Stanford University · Nov 2026
- Independenteess.IV
X-ray AI can describe a collapsed lung while missing every case
Medical vision-language models can appear accurate on average while missing every case of a critical diagnosis.
Independent Researcher · Nov 2026
- Academiccs.MA
Adaptive interaction routing improves multimodal AI debate accuracy while cutting token use
AIM boosts multimodal debate accuracy by adaptively choosing text, graph, or region-based interaction per query.
Virginia Tech, Inha University · Nov 2026
- Big Techcs.AI
Language models can learn financial labels better without teacher explanations
Response-only LoRA fine-tuning improves domain performance while preserving general capabilities better than incorporating reasoning traces.
NVIDIA · Nov 2026
- Academiccs.CR
Language models can conceal watermarked training data by altering outputs
Logit Laundering reshapes inference-time logits to remove detectable watermark signals, enabling evasion of data-use auditing without model retraining.
Beijing University of Posts and Telecommunications, Zhongguancun Laboratory · Nov 2026
- Big Techcs.CV
AI models forecast satellite changes better without predicting individual pixels
Forecasting future satellite observations as foundation-model embeddings instead of pixel spectra halves prediction error and gives far better-calibrated uncertainty.
University of Cambridge, Microsoft AI for Good Research Lab · Nov 2026
- Top Universitycs.LG
Stochastic latent variable improves masked image prediction without hurting representations
Making the predictor's hidden variable stochastic in Joint-Embedding Predictive Architectures improves representation quality and adds uncertainty estimates and diverse predictions.
Imperial College London, Information Technologies Institute, CERTH · Nov 2026
- Research Labcs.LG
Distillation under-weights the tokens a model is most confident about
Standard on-policy distillation starves high-confidence tokens of gradient; matching teacher and student probabilities directly fixes that and improves reasoning scores.
University of Surrey, University College London · Nov 2026
- Top Universitycs.AI
Fast AI agents read states correctly yet still choose wrong actions
Fast agent models mostly read state correctly but still act wrongly, due to compositional and sub-goal failures.
Carnegie Mellon University · Oct 2026
- Top Universitycs.AI
AI agents halve simulated harmful outcomes when safeguards weigh consequences
Allocating scrutiny by consequence-weighted risk and reversibility roughly halves harmful side effects of tool-using agents compared with uncertainty ranking, in controlled simulation.
Carnegie Mellon University, Feinity LLC · Oct 2026
- Academiccs.CR
Jev Resists Rule-Based Attacks but Falters on Phishing Judgments
Jev resists adversarial forgeries on rule-based moderation but is less robust and less accurate on judgment-based phishing detection.
Nankai University, Zhongguancun Laboratory · Oct 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
- Chinese Techcs.CL
Medical decision models stay consistent but struggle to abstain reliably
Jev matches leading models on text-heavy medical tasks and resists format changes, but struggles with numerical decisions and does not abstain reliably.
Harbin Institute of Technology, Ant Group · Oct 2026
- Big Techcs.CL
Alternating context edits and logit steering boosts frozen LLM accuracy
EMO improves frozen language-model reasoning by alternating judge-guided context revision with confidence-constrained steering derived from context-induced logit changes.
Peking University, Samsung Research · Oct 2026
- Research Labcs.AIcode
Agent performance depends more on system configuration than model size
Agent performance depends more on system configuration, especially supplied information, than on model choice alone, and reliable verification requires system tools rather than prompts.
Technical University of Munich, MCML · Oct 2026
- Top Universitycs.AI
Looped decision models spend extra computation on harder questions
LEV lets a decision model improve accuracy on hard questions by looping its layers, with a stopper saving computation per question.
Lehigh University, University of British Columbia · Oct 2026
- Top Universitycs.AI
Open language models match or beat a proprietary decision model without specialized training
Open instruction-tuned LLMs can match Jev through answer-letter logits and surpass it after brief supervised LoRA fine-tuning.
Harvard University, Kempner Institute · Oct 2026
- Big Techcs.CL
Reconstructing text under safety labels produces more reliable guard models
LLaDA-Guard scores text under each safety label hypothesis and classifies based on which label better explains the content.
University of Neuchâtel, Delft University of Technology · Sept 2026
- Chinese Techcs.CL
Multimodal decision models output structured choices in a single pass
Intern-Decision makes structured multimodal decisions in a single forward pass, achieving high accuracy and low latency.
Shanghai AI Lab, Shanghai Jiao Tong University · Oct 2026
- Chinese Techcs.LG
Bayesian belief framework jointly handles two uncertainties in process reward models
BayesianPRM improves process reward modeling by jointly handling PRM uncertainty and policy uncertainty with reliability- and conservatism-aware Bayesian belief updates.
CityUHK, Shanghai Innovation Institute · Oct 2026
- Chinese Techcs.CL
LLMs often diagnose without correctly valuing available medical evidence
LLMs often submit correct diagnoses on insufficient or misleading evidence, and verifying evidential support before submission improves diagnostic accuracy.
Zhejiang University, Ant Healthcare, Ant Group · Sept 2026
- Top Universitycs.CR
Unverified state text can flip calibrated model decisions
Appending an unverified opinion to the state flips 12.1% of decisions in RLCD models, making the state an untrusted input.
Institute of Information Engineering, Chinese Academy of Sciences, University of Chinese Academy of Sciences · Sept 2026
- Top Universitycs.AI
Bayesian reliability memory helps agents decide when to trust advice
BaRe-Mem uses Bayesian updates from past interactions to estimate advisor reliability, which then guides consultation decisions and response weighting.
DeCLaRe Lab, Nanyang Technological University, Peking University · Sept 2026
- Academiccs.AI
Neither decision models nor new LLMs beat statistical baselines as synthetic survey respondents
Decision models produce plausible survey marginals but lack psychometric validity, while newer generative models do not improve on older models or statistical baselines.
Hostinger, Kaunas University of Technology · Oct 2026
- Independentcs.LG
Fixed LLM judges can mismeasure upgrades as agents change
Fixed LLM judges make version-dependent errors when evaluating agent upgrades, making paired audits preferable.
Sept 2026
- Academiccs.CL
Typed decision models show bias but offer cheaper, consistent judgment
All typed decision models tested are biased: emotional tone can cut a typical judge's accuracy from 99% to 42%, and they are more biased than 2026 LLM judges.
University of Göttingen, Bielefeld University of Applied Sciences and Arts (HSBI) · Oct 2026
- Chinese Techcs.SD
Direct decision head makes bounded audio judgments about 20 times faster
Qwen-Audio-Jev adds a dedicated decision head to an audio language model, directly outputting choices, binary judgments, and scores, with a 20x inference speedup.
Token Foundry, Alibaba Group · Oct 2026
- Big Techcs.LG
A synthetic-trained model predicts across tabular datasets without retraining
TabFM is a zero-shot foundation model for tabular data that outperforms tuned AutoML pipelines across 51 benchmarks.
Google Research · Oct 2026
- Big Techcs.CL
Adapting a speech language model with self-generated prefix supervision improves simultaneous translation quality-latency trade-offs.
Prefix learning from partial speech improves quality-latency trade-offs in simultaneous translation, especially with multi-turn decoding and confidence thresholds.
Microsoft · Oct 2026
- Academiccs.OS
Prioritizing action-relevant context cuts LLM agent cache costs
ActKV retains 98.53% of FullKV accuracy using 25.98% peak KV cache memory, achieving 3.97x token and 3.58x task throughput.
University of Science and Technology of China, Suzhou Institute for Advanced Research, University of Science and Technology of China · Sept 2026
- Academiccs.CL
Batching makes AI grading verdicts depend on the other items
Batching multiple items in one LLM prompt makes verdicts on ambiguous items depend on which other items and where they appear.
Technical University of Munich · Oct 2026
- Research Labcs.CL
Task-aware certificates safely automate part of LLM agent evaluation
A task-level bootstrap certificate ensures LLM judge error stays under budget, enabling safe selective automation of agent evaluation.
Independent Researcher, University of Chinese Academy of Sciences · Sept 2026
- Academiccs.CL
Short natural context can flip decision models to wrong answers
Short, natural, answer-preserving context additions can redirect decision models from correct choices to specific high-confidence wrong choices.
University of Southern California · Sept 2026
- Big Techcs.CL
Confidence-guided complementary evidence pool improves multimodal retrieval-augmented generation
CLIMB selects complementary evidence and uses confidence scores to decide when to update answers, improving multimodal RAG without retraining.
Rutgers University, Google · Oct 2026
- Research Labcs.DC
Adaptive verification speeds block-diffusion decoding under heavy concurrent demand
DScale improves throughput of block-diffusion speculative decoding by up to 48.8% over DFlash with only a 112K-parameter predictor.
Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, University of Chinese Academy of Sciences · Oct 2026
- Big Techcs.AI
On-device caches convert natural language to actions with classification not generation
On-device operation caches reduce latency, cost, and privacy risks by converting NL-to-Action generation into a classification problem.
University of Kentucky, Microsoft · Sept 2026
Concepts are extracted from each paper and reused across the corpus, so this page grows on its own as the desk reads.