CISPO
PulseAugur coverage of CISPO — every cluster mentioning CISPO across labs, papers, and developer communities, ranked by signal.
1 day(s) with sentiment data
-
Nvidia details Nemotron 3.5 Super RL post-training gains
Nvidia has detailed its Nemotron 3.5 Super model's performance improvements after post-training with reinforcement learning on expert-built tasks. The model, specifically its 120B parameter version with a 12B active par…
-
New Proximal KL method improves off-policy RLVR accuracy on Qwen3 models
Researchers have developed Proximal KL, a novel method for off-policy reinforcement learning with verifiable rewards (RLVR). This technique addresses challenges like asynchronous generation and response reuse by using t…
-
27B LLM trained to generate profitable trading formulas
A research paper details the development of a 27 billion parameter model, Qwen3.8-27B, trained using Multi-Reward Reinforcement Learning to generate one-line trading formulas. These formulas are then converted into dail…
-
LLM RL techniques boost Qwen3.8-27B scores, but an "advantage floor" hinders learning
Researchers have explored advanced reinforcement learning techniques, specifically focusing on multi-reward optimization for large language models. In their latest experiments with Qwen3.8-27B, they found that combining…
-
Qwen3 14B model experiments compare reinforcement learning techniques · 2 sources tracked
Experiments were conducted on the Qwen3 14B model to evaluate various reinforcement learning techniques, including GRPO, DAPO++, CISPO, Adapoides, and REPO-RECK. The study focused on comparing these methods with and wit…
-
Mistral AI releases Leanstral 1.5 for advanced formal verification
Mistral AI has released Leanstral 1.5, an open-source model designed for formal verification tasks. This model, which has 6 billion active parameters and is available under the Apache 2.0 license, demonstrates significa…
-
New RLVR method ACPO enhances LLM reasoning capabilities
Researchers have analyzed Reinforcement Learning from Verifiable Rewards (RLVR) to understand its impact on large language model reasoning. Their theoretical analysis revealed that the degree of off-policy learning, inf…