SmolLM-135M-Instruct
PulseAugur coverage of SmolLM-135M-Instruct — every cluster mentioning SmolLM-135M-Instruct across labs, papers, and developer communities, ranked by signal.
2 day(s) with sentiment data
-
New W-ISA system enables reproducible weight-space interventions in language models
Researchers have introduced W-ISA, a new software instruction set and runtime designed for executing weight-space intervention programs on pre-trained language models. This system allows for precise parameter updates, p…
-
LLMs Learn to Read and Write Neural Network Weights Directly
Researchers are exploring "Weight-Space Programming" (WSP), a novel technique where Large Language Models (LLMs) can directly read, diagnose, and modify the raw numerical weights of neural networks. This approach bypass…
-
Developer trains 75M parameter LLM that outperforms larger models
A developer has trained a 75 million parameter language model called KeyLM from scratch, utilizing 18 billion tokens for pre-training. The instruction-tuned version of KeyLM demonstrates superior performance on the IFEv…