Update (base update)

[ghstack-poisoned]

Author: vmoens

docs/source/reference/llms.rst

@@ -1155,20 +1155,22 @@ Objectives
 
 LLM post-training requires specialized loss functions that are adapted to the unique characteristics of language models.
 
-GRPO
-~~~~
-
-The :class:`~torchrl.objectives.llm.GRPOLoss` class is a thin wrapper around the :class:`~torchrl.objectives.PPOLoss` class
-that codes the LLM-specific functionalities.
+GRPO, DAPO, CISPO
+^^^^^^^^^^^^^^^^^
 
 .. currentmodule:: torchrl.objectives.llm
 
 .. autosummary::
     :toctree: generated/
     :template: rl_template.rst
 
+    LLMLossOutput
     GRPOLoss
     GRPOLossOutput
+    CISPOLoss
+    CISPOLossOutput
+    DAPO
+    DAPOLossOutput
     MCAdvantage
 
 SFT