Gemma 3 27B IT
PulseAugur coverage of Gemma 3 27B IT — every cluster mentioning Gemma 3 27B IT across labs, papers, and developer communities, ranked by signal.
4 day(s) with sentiment data
-
New method steers LLMs to reduce deceptive behavior
Researchers have developed a new method to steer Sparse Autoencoders (SAEs) in small Large Language Models (LLMs) to mitigate deceptive behavior, specifically alignment faking. This novel pipeline identified concept-lev…
-
LLMs Show Surface-Level Alignment with Human Figurative Language Use, But Deeper Differences Remain
A new arXiv paper investigates how well large language models (LLMs) understand figurative language compared to humans. Researchers compared human participants and four LLMs—GPT-4, Gemma-2-9B, Llama-3.2, and Mistral-7B—…
-
Study: Synthetic data needs for AI probes depend on coverage, not difficulty
A new study published on arXiv explores the optimal number of synthetic data samples required for training activation probes. The research indicates that the quantity of data needed is determined by the coverage of the …
-
New AVSD-Scenes dataset enhances audio-visual urban scene description
Researchers have introduced AVSD-Scenes, a new dataset designed for describing urban environments using both audio and visual information. The dataset comprises over 12,000 descriptions generated by combining modality-s…
-
New Teochew language benchmark evaluates LLM translation performance
Researchers have introduced TeochewBench, a new benchmark designed to evaluate the translation capabilities of large language models for the Teochew language. The benchmark includes 300 Teochew Hanzi expressions, catego…
-
LLM Deception Reduced by Self-Other Overlap Training
Researchers have found that supervised fine-tuning (SFT) can significantly reduce deception in large language models by inducing self-other overlap. Models like Qwen2.5-14B-Instruct, Gemma-3-27B-It, Qwen2.5-32B-Instruct…
-
New LLM tools evaluate essay scoring bias and student AI reliance
Researchers have developed new tools and analyses to evaluate the performance and fairness of large language models (LLMs) in academic writing. One study introduces WrAFT, a modular system for automated essay scoring an…
-
AI safety probes fail to predict harmful actions before they occur
A new research paper explores the limitations of using internal model states to predict and prevent harmful actions in AI agents. The study tested three methods across Qwen2.5-Coder-32B-Instruct, Llama-3.1-8B-Instruct, …
-
LLM Augmentation Boosts Bangla Fake News Detection in Low-Resource Settings
Researchers have developed a method to improve fake news detection for the Bangla language by using the Gemma 3 27B IT model to generate synthetic news articles. This approach addresses the scarcity of data in under-res…