Research subjects A–Z
Every subject in Life of Research, drawn from the keywords authors give their work. Each subject page lists all of its open-access papers, preprints and datasets, newest first.
A
- Ablation 2
- Abstention 4
- Abstention policy 1
- Acceptance tests 1
- Activation patching 1
- Activation probes 1
- Activation steering 1
- Active learning 1
- Adaptive attacks 1
- Adversarial robustness 1
- Agent evaluation 4
- Agent harness 1
- Agent memory 12
- Agent protocols 1
- Agent safety 3
- AI agents 1
- AI control 2
- AI evaluation 1
- AI-generated text 1
- AI governance 1
- AI safety 2
- AI security 2
- Alignment auditing 1
- Ambiguity 1
- Analogical reasoning 1
- Annotation completeness 1
- Annotation cost 1
- Anomaly detection 3
- Answerability 1
- API auditing 1
- Artificial immune systems 1
- Astrophysique théorique 1
- Attack success rate 1
- Attribution 1
- Autoformalization 1
- Automated scientific discovery 1
- Automatic curriculum learning 1
- Automation bias 1
- Autorater validity 1
- Autotelic agents 1
- Auxiliary task demands 1
B
C
- Calibration 7
- Capability estimation 2
- Capability forecasting 2
- Cascade containment 2
- Catastrophic forgetting 3
- Category discovery 2
- Causal abstraction 2
- Causal intervention 1
- Causal mediation analysis 1
- Chain-of-thought 1
- Chain-of-thought faithfulness 2
- Chain-of-thought monitoring 2
- Chain-of-thought reasoning 2
- Chain-of-thought verification 1
- Change detection 1
- Circuit discovery 1
- Clarification questions 1
- Class-agnostic proposals 1
- Class number estimation 1
- Clinical artificial intelligence 1
- Clustering 1
- Code execution 1
- Code signing 1
- Collective intelligence 2
- Colonoscopy 1
- Compensation 1
- Complementary learning systems 1
- Compositional risk 1
- Compute-matched baselines 1
- Compute-normalised evaluation 1
- Computer-aided detection 1
- Concept drift 2
- Concept evolution 1
- Confidence calibration 1
- Conflict resolution 1
- Conformal prediction 1
- Conformity 1
- Confounding 1
- Construct validity 13
- Context-aware safety 1
- Context faithfulness 1
- Context-memory conflict 1
- Contextual integrity 1
- Continual learning 4
- Copyright 1
- Correlated errors 4
- Corrigibility 1
- Cost-aware evaluation 1
- Cost-controlled evaluation 1
- Counterfactual intervention 1
- Counterfactual reasoning 1
- Counterfactual tasks 1
- Credit assignment 1
- Criterion validity 1
- Cross-domain generalisation 2
- Curiosity 2
- Curriculum learning 1
D
- Danger theory 1
- Dangerous capability evaluations 1
- DAPO 1
- Data attribution 1
- Data-constrained scaling 1
- Data contamination 1
- Data contamination detection 1
- Data curation 1
- Data deletion 1
- Data poisoning 2
- Data provenance 2
- Data repetition 1
- Data selection 1
- Data streams 1
- Data valuation 1
- Deception detection 1
- Declassification 2
- Decontamination 1
- Deduplication 1
- Delegation 2
- Deletion semantics 2
- Dependability 1
- Deserialization 1
- Deskilling 1
- Dictionary learning 1
- Difficulty filtering 1
- Distribution shift 3
- Dual-process theory 2
E
- Early warning 1
- Emergent abilities 2
- Employee Benefits Kuwait 1
- End of Service Indemnity 1
- Ensemble diversity 1
- Enthymeme 1
- Epistemic uncertainty 1
- Error correlation 1
- Error propagation 3
- Error slice discovery 2
- Estimated performance gain 1
- EU AI Act 1
- Evaluation 9
- Evaluation methodology 13
- Evaluation metrics 2
- Evaluation protocol 2
- Evaluation science 2
- Evaluation validity 4
- Evaluator validity 1
- Evidence sufficiency 1
- Exchangeability 1
- Experiential learning 1
- Experimental design 1
- Explainability 1
- Explainable anomaly detection 1
- Explanation faithfulness 1
- Exploration 2
- Exploration bonus 2
- Extreme verification latency 1
F
G
H
I
- Identifiability 4
- Implicit premises 1
- Importance scoring 1
- Information-flow control 5
- Instruction following 2
- Instruction hierarchy 1
- Integrity 2
- Inter-annotator agreement 1
- Inter-rater agreement 1
- Internal contradiction 1
- Interpretability 1
- Intrinsic motivation 3
- Introspection 3
- Inverse scaling 2
- Irreversibility 1
J
K
L
- Label granularity 2
- Label preservation 1
- Language model evaluation 1
- Language model pretraining 1
- Language models 1
- Large audio-language models 1
- Large language model agents 2
- Large language models 8
- Learning progress 1
- Learning to defer 2
- Least privilege 1
- Letter-string analogies 1
- Linear representation hypothesis 1
- Literature-based discovery 1
- LLM agent safety 2
- LLM agents 10
- LLM-as-a-judge 8
- LLM auditing 1
- LLM collectives 1
- LLM evaluation 2
- LLM reasoning 2
- LLM routing 2
- LLM security 1
- LLM serving 1
- Localization 1
- Long-context inference 1
- Long-horizon execution 1
- Long-term conversational memory 1
- Long-term memory 1
- Loss of plasticity 1
M
- Machine learning supply chain 1
- Machine self-awareness 1
- Machine unlearning 4
- Majority voting 1
- Mammography screening 1
- Measurement 1
- Measurement theory 1
- Mechanistic interpretability 4
- Membership inference 3
- Memorization 2
- Memory admission control 1
- Memory consolidation 1
- Memory evaluation 2
- Memory eviction 1
- Memory poisoning 4
- Memory retrieval 1
- Meta 1
- Metacognition 3
- Metacognitive control 3
- Modality reliance 1
- Model collapse 2
- Model compression 1
- Model Context Protocol 1
- Model drift 1
- Model editing 1
- Model fingerprinting 1
- Model heterogeneity 2
- Model hubs 1
- Model lineage 1
- Model merging 1
- Model monitoring 1
- Model self-knowledge 2
- Model substitution 1
- Monte Carlo estimation 2
- Multi-agent debate 2
- Multi-agent systems 8
- Multi-party conversation 1
- Multimodal evaluation 1
- Multitrait-multimethod 1
N
O
- Open-ended learning 1
- Open-set active learning 1
- Open-set label shift 1
- Open-set recognition 1
- Open-set test-time adaptation 1
- Open-vocabulary recognition 1
- Open-world learning 2
- Open-world object detection 1
- Operating point 1
- Opt-out 1
- Organizational learning 1
- Out-of-distribution data 1
- Out-of-distribution detection 5
- Over-refusal 1
- Overthinking 2
- Ownership verification 1
P
- Parameter resets 1
- Parametric knowledge 1
- Parametric memory 1
- Paraphrase robustness 1
- Partial-input baselines 1
- Pass@k 2
- Peer review 1
- Permission disclosure 1
- Persona prompting 2
- Persona stability 1
- Plan-level safety 1
- Poison detection 1
- Policy conflict 1
- Post-training 2
- Post-training quantization 1
- Pretraining data 2
- Prioritized experience replay 1
- Privacy 3
- Privilege ordering 1
- Process reward models 2
- Process supervision 3
- Prompt compression 1
- Prompt injection 6
- Proof-of-learning 1
- Provenance 1
- Psychometrics 1
Q
R
- Rate-distortion 1
- Reasoning 2
- Reasoning boundary 2
- Reasoning models 3
- Recency ranking 1
- Recursive self-modification 1
- Refusal 1
- Refusal calibration 1
- Refusal direction 1
- Reinforcement learning with verifiable rewards 8
- Relearning attacks 1
- Rephrasing 1
- Representation engineering 1
- Reproducibility 3
- Reputation systems 1
- Research idea generation 1
- Resource awareness 1
- Responsible scaling 1
- Retention policy 1
- Retrieval-augmented generation 6
- Retrieval-augmented verification 1
- Reward hacking 1
- Reward model evaluation 1
- Right to be forgotten 1
- Right to erasure 2
- RLVR 3
- Robustness 1
- Root cause analysis 2
- Runtime enforcement 1
S
- Safe completion 1
- Safety alignment 1
- Safety benchmarks 1
- Sarcasm 1
- Scaffolding 1
- Scalable oversight 3
- Scheming 1
- Science of science 1
- Scientometrics 1
- Security evaluation 1
- Selective answering 1
- Selective classification 1
- Selective prediction 1
- Self-bias 1
- Self-consistency 2
- Self-correction 4
- Self-explanation faithfulness 1
- Self-improvement 1
- Self-improving agents 1
- Self-prediction 1
- Self-repair 1
- Self-report 2
- Self-training 1
- Semantic duplicates 1
- Semantic shift 1
- Sentiment analysis 1
- Sequential monitoring 1
- Shared memory 1
- Shnoudi Theory.. Mathematics..Physics 1
- Shortcut learning 1
- Shutdown problem 1
- Skill retention 1
- Social bias 1
- Source credibility 1
- Sparse autoencoders 1
- Spoken dialogue systems 1
- Stability-plasticity 1
- Staleness 1
- Statistical testing 1
- Steering vectors 1
- Step-level verification 2
- Study design 1
- Summarization faithfulness 1
- Superposition 1
- Survival analysis 2
- Synthetic data 2
T
- Task affinity 2
- Task decomposition 1
- Temporal knowledge 1
- Temporal reasoning 1
- Test-time adaptation 1
- Test-time compute 2
- Test-time compute allocation 1
- Test-time scaling 5
- Text priors 1
- Threat modelling 3
- Time horizon 1
- Token eviction 1
- Tool hallucination 1
- Tool use 1
- Training data provenance 1
- Transfer 1
- Transfer learning 2
- Trust modeling 1
- Trusted monitoring 1
- Turn-taking 1
U
V
W
No subject matches that filter. Try a full-text search.