SAC
E426679
SAC (Soft Actor-Critic) is a popular off-policy deep reinforcement learning algorithm that optimizes both expected return and policy entropy to achieve stable and efficient learning in continuous control tasks.
All labels observed (1)
| Label | Occurrences |
|---|---|
| SAC canonical | 2 |
How this entity was disambiguated
This entity first appeared as the object of triple T4277524 — resolving that mention is where its identity was fixed. The disambiguator weighed these candidate entities and picked the highlighted one (or “None”, minting a new entity). This is how homonymy is resolved: the same surface form can point to different entities.
NED1
Entity disambiguation (via context triple)
gpt-5-mini-2025-08-07
Target entity: SAC Context triple: [TF-Agents, supportsAlgorithmFamily, SAC]
-
A.
SAC
SAC is a NATO-led multinational program that provides participating nations with shared strategic airlift capabilities using a fleet of C-17 Globemaster III aircraft.
-
B.
SACT
SACT is the Supreme Allied Commander Transformation, the NATO strategic commander responsible for leading the alliance’s military transformation and capability development.
-
C.
SACS
SACS is a regional accrediting body in the United States that evaluates and certifies the quality and standards of educational institutions in the southern states.
-
D.
SAC Steering Board
The SAC Steering Board is the multinational governing body that oversees and directs NATO’s Strategic Airlift Capability program on behalf of its participating nations.
-
E.
SacRT
SacRT is the public transit agency serving the Sacramento, California metropolitan area with bus, light rail, and related transportation services.
- F. None of above. chosen
- G. Unsure - the case is ambiguous/there is not enough information to decide.
NED2
Entity disambiguation (via description)
gpt-5-mini-2025-08-07
Target entity: SAC Target entity description: SAC (Soft Actor-Critic) is a popular off-policy deep reinforcement learning algorithm that optimizes both expected return and policy entropy to achieve stable and efficient learning in continuous control tasks.
-
A.
SAC
SAC is a NATO-led multinational program that provides participating nations with shared strategic airlift capabilities using a fleet of C-17 Globemaster III aircraft.
-
B.
SACT
SACT is the Supreme Allied Commander Transformation, the NATO strategic commander responsible for leading the alliance’s military transformation and capability development.
-
C.
SACS
SACS is a regional accrediting body in the United States that evaluates and certifies the quality and standards of educational institutions in the southern states.
-
D.
SAC Steering Board
The SAC Steering Board is the multinational governing body that oversees and directs NATO’s Strategic Airlift Capability program on behalf of its participating nations.
-
E.
SacRT
SacRT is the public transit agency serving the Sacramento, California metropolitan area with bus, light rail, and related transportation services.
- F. None of above. chosen
Statements (49)
| Predicate | Object |
|---|---|
| instanceOf |
deep reinforcement learning algorithm
ⓘ
deep reinforcement learning algorithm ⓘ off-policy reinforcement learning algorithm ⓘ off-policy reinforcement learning algorithm ⓘ |
| abbreviation | SAC NERFINISHED ⓘ |
| advantageOverDeterministicMethods | better exploration via entropy maximization ⓘ |
| aimsFor |
sample-efficient learning
ⓘ
stable learning ⓘ |
| canBe | model-free ⓘ |
| category | actor-critic methods ⓘ |
| commonlyEvaluatedOn |
MuJoCo benchmarks
NERFINISHED
ⓘ
OpenAI Gym continuous control tasks ⓘ |
| comparedWith |
DDPG
NERFINISHED
ⓘ
TD3 NERFINISHED ⓘ |
| criticUpdateBasedOn | soft Bellman backup ⓘ |
| firstPublishedIn | "Soft Actor-Critic: Off-Policy Maximum Entropy Deep Reinforcement Learning with a Stochastic Actor" NERFINISHED ⓘ |
| fullName | Soft Actor-Critic NERFINISHED ⓘ |
| hasVariant |
automatic entropy tuning SAC
ⓘ
discrete SAC ⓘ |
| implementedIn |
PyTorch
NERFINISHED
ⓘ
TensorFlow NERFINISHED ⓘ |
| introducedIn | 2018 ⓘ |
| is | maximum entropy reinforcement learning method ⓘ |
| laterExtendedIn | "Soft Actor-Critic Algorithms and Applications" NERFINISHED ⓘ |
| learningSignal | temporal-difference error ⓘ |
| objectiveIncludes | temperature parameter ⓘ |
| optimizationObjective |
expected return
ⓘ
policy entropy ⓘ |
| policyOutput | distribution over continuous actions ⓘ |
| policyType | stochastic policy ⓘ |
| policyUpdateBasedOn | reparameterization trick ⓘ |
| proposedBy |
Aurick Zhou
NERFINISHED
ⓘ
Pieter Abbeel NERFINISHED ⓘ Sergey Levine NERFINISHED ⓘ Tuomas Haarnoja NERFINISHED ⓘ |
| sampleEfficiency | high ⓘ |
| supports | continuous action spaces ⓘ |
| temperatureParameterControls | entropy-returns trade-off ⓘ |
| trainingStability | high ⓘ |
| typicalDomain | continuous control tasks ⓘ |
| updateStyle | off-policy updates ⓘ |
| usedIn |
autonomous driving research
ⓘ
manipulation tasks ⓘ robotics control ⓘ |
| uses |
actor-critic architecture
ⓘ
entropy regularization ⓘ replay buffer ⓘ target networks ⓘ |
| valueFunctionType | soft Q-function ⓘ |
How these facts were elicited
The pipeline generated the facts above by prompting gpt-5.1 with this entity's name + description and the instruction below.
Instruction
You are a knowledge base construction expert. Given a subject entity and a description of it, return factual statements that you know for the subject as a JSON list of dictionaries(triples), where keys must be "subject", "predicate" and "object". The number of facts may be very high, between 25 to 50 or more, for very popular subjects. For less popular subjects, the number of facts can be very low, like 5 or 10. # Requirements - If you don't know the subject at all, return an empty list. - If the subject is not a named entity, return an empty list. - Include at least one triple where predicate is "instanceOf". - Do not get too wordy. - Separate several objects into multiple triples with one object.
Input
Subject: SAC Description of subject: SAC (Soft Actor-Critic) is a popular off-policy deep reinforcement learning algorithm that optimizes both expected return and policy entropy to achieve stable and efficient learning in continuous control tasks.
Referenced by (2)
Full triples — surface form annotated when it differs from this entity's canonical label.