Emergent Ability
Emergent Ability refers to the phenomenon in which complex systems suddenly exhibit new capabilities or behavioral characteristics not present at smaller scales once a critical threshold of scale or complexity is reached. In artificial intelligence, the term specifically denotes capabilities such as reasoning, analogy, and multi-step problem solving that spontaneously appear in large language models once parameter count and training data volume surpass a certain threshold. A defining characteristic is nonlinearity and unpredictability: the emergence of capabilities manifests as a phase transition rather than a gradual improvement.
Core Facts
When a model's parameter count surpasses a certain critical threshold, capabilities such as reasoning and debugging that are not seen in smaller models emerge, a phenomenon known as Emergent Abilities.
A 2023 Stanford University study challenged the phenomenon of "emergence," arguing that it is largely an artifact of the choice of evaluation metrics rather than a genuine sudden shift in model capability.
Frontier models exhibit emergent capabilities, meaning they suddenly display abilities not explicitly included in the training objective once a certain parameter-scale threshold is reached.
Timeline (last 90 days)
When multiple AI agents interact, the system as a whole may produce collective behaviors that a single agent does not possess, such as coordinated probing, task allocation and strategy evolution.
Emergent capabilities are new abilities that suddenly appear once an AI system's scale or complexity exceeds a certain threshold and that are not explicitly programmed or directly specified by the training objective.
The "emergent abilities" required for complex reasoning usually appear only in models with larger parameter counts.
Emergent capabilities refer to reasoning, code generation, or tool-use abilities that models spontaneously exhibit beyond their training objectives.
The emergent capabilities of large language models are regarded by some researchers as signals of general intelligence, while others believe these capabilities may still represent efficient compression and retrieval of implicit patterns in training data.
Emergent capabilities create structural limitations in the accuracy of ex ante risk assessment and challenge the logic of foreseeability determination in traditional product liability law
Current large language models lack reliable self-introspection capabilities.
The industry divides model capabilities into two dimensions: emergent capabilities and alignment
The "capability risk" camp, represented by Geoffrey Hinton and Yoshua Bengio, believes that current large models have already exhibited emergent capabilities that are difficult to fully explain, and that engineering measures alone cannot eliminate the fundamental risk of alignment failure.
A 2023 Stanford University study pointed out that some abilities considered emergent may be artifacts of evaluation metric selection rather than genuine qualitative changes in model capability.
19 more timeline events
All Facts (20)
When a model's parameter count surpasses a certain critical threshold, capabilities such as reasoning and debugging that are not seen in smaller models emerge, a phenomenon known as Emergent Abilities.
90%VerifiedA 2023 Stanford University study challenged the phenomenon of "emergence," arguing that it is largely an artifact of the choice of evaluation metrics rather than a genuine sudden shift in model capability.
85%VerifiedFrontier models exhibit emergent capabilities, meaning they suddenly display abilities not explicitly included in the training objective once a certain parameter-scale threshold is reached.
80%VerifiedEmergent ability refers to the phenomenon where, once a model's scale exceeds a certain critical threshold, certain previously absent capabilities suddenly appear and performance undergoes a sharp leap.
80%VerifiedA 2023 Stanford paper argued that emergence is a measurement artifact caused by the choice of evaluation metrics, and that with smooth metrics the capability improvement is continuous.
80%VerifiedThe emergent capability of large language models refers to the complex reasoning ability that suddenly manifests after the parameter scale surpasses a certain threshold.
80%VerifiedIn 2022, the Google Brain team first systematically documented the phenomenon of 'emergent abilities' in large language models in the paper 'Emergent Abilities of Large Language Models'
80%VerifiedEmergent abilities refer to capabilities that appear suddenly when a model's scale exceeds a certain threshold and are absent in smaller-scale models, such as chain-of-thought reasoning.
75%Verified涌现能力（emergent abilities）指模型规模超过某临界点时突然显现训练时未明确优化的新能力，如多步推理、代码生成或少样本学习
75%VerifiedThe chain-of-thought effect is particularly pronounced in models with parameter counts exceeding 100B.
75%VerifiedRylan Schaeffer's team at Stanford University published a rebuttal study in 2023 arguing that "emergence" may be a visual illusion caused by the discontinuous selection of evaluation metrics, rather than a true nonlinear leap in model capability
70%VerifiedAI red-teaming faces the special challenge of emergent capabilities, as models may exhibit capabilities under specific prompt combinations that were not anticipated during training, making exhaustive testing nearly impossible
70%VerifiedCurrent large language models lack reliable self-introspection capabilities.
65%VerifiedLarge language models exhibit emergent capabilities, suddenly displaying abilities not present at smaller scales once the model size exceeds a certain threshold.
65%VerifiedThe mechanism behind the emergence of emergent capabilities remains a hotly debated topic in academia, with some researchers believing it is a genuine phase-transition phenomenon and others believing it is a visual artifact caused by evaluation metrics
65%VerifiedSchaeffer et al. from Stanford questioned the emergence phenomenon, arguing that emergence may partly be a measurement artifact caused by the choice of evaluation metrics.
65%VerifiedA 2022 Google research report was the first to systematically document the phenomenon of emergent abilities.
65%Unverified部分研究者认为大模型涌现能力是真实的相变现象，另一些人认为这只是评估指标选择造成的统计假象
85%UnverifiedThe emergent capabilities demonstrated by large language models make technological progress highly unpredictable, as models suddenly exhibit new capabilities that their trainers never explicitly designed when model scale crosses a certain computational threshold.
70%UnverifiedEmergent Abilities refer to capabilities that large language models suddenly exhibit after the scale or training data reaches a certain threshold, which they did not previously possess.
60%