The Twelfth International Conference on Learning Representations
PulseAugur coverage of The Twelfth International Conference on Learning Representations — every cluster mentioning The Twelfth International Conference on Learning Representations across labs, papers, and developer communities, ranked by signal.
1 day(s) with sentiment data
-
New SalUn technique enables AI models to forget specific data
A new machine unlearning technique called SalUn, presented at ICLR 2024, allows AI models to forget specific data by targeting and altering the most relevant weights. This method achieves high unlearning accuracy on dat…
-
Local LLM measurements reveal issues with 'thinking' and self-critique
Running local LLMs like Qwen3.5 4B requires careful measurement and verification, as results can be inconsistent. Enabling 'thinking' in some models can lead to empty responses if the token budget is exceeded, necessita…
-
Open-source AI agent uses calibration and external feedback for trading
This article details the technical mechanisms behind an open-source AI agent system designed to approach a "world model" for trading. The system employs a predict-act-verify loop, with a strong emphasis on external, qua…
-
AI watermarks can be detected and persist through model training chains
Researchers have developed methods to watermark AI-generated text, making it statistically detectable. This technique, pioneered by Kirchenbauer et al. and validated by Google's SynthID-Text within Gemini, embeds a subt…
-
Verify coding agent reports, not just their output, developer advises
A software developer highlights the critical need to verify the output of coding agents, rather than trusting their self-reported success claims. The developer recounts instances where agents confidently reported succes…