counterfactual history
PulseAugur coverage of counterfactual history — every cluster mentioning counterfactual history across labs, papers, and developer communities, ranked by signal.
1 day(s) with sentiment data
-
LadderEdit cuts LLM editing memory by 5.2x with residual compression · arXiv
Researchers have developed LadderEdit, a novel method for memory-efficient lifelong editing of Large Language Models (LLMs). This approach compresses individual LoRA adapters after each edit is acquired, initially stori…
-
AI knowledge-editing benchmarks flawed, new study finds
A new research paper argues that current benchmarks for knowledge-editing in AI models are fundamentally flawed, failing to accurately measure the crucial "scope classification" decision. The authors developed a gradien…
-
New framework RARE improves MoE language model steering
Researchers have developed RARE, a new framework for steering Mixture-of-Experts (MoE) language models, which decouples representation steering from expert routing. This approach projects behavioral perturbations onto t…
-
PRISM Edit enhances LLM temporal fact updates without retraining
Researchers have developed PRISM Edit, a novel method for updating temporal facts in large language models without full retraining. Unlike traditional methods that replace information, PRISM Edit optimizes a single repr…
-
New MAAT unlearning method tackles 'Why' questions with balanced benchmark
Researchers have introduced MAAT, a novel three-phase framework for targeted machine unlearning that specifically addresses the difficulty of removing causal knowledge. Existing benchmarks are skewed, underrepresenting …