Most parameters of a Mixture-of-Experts (MoE) model belong to its experts, and standard training keeps every expert gradient in GPU memory until the optimizer step. This memory grows with the number and size of the experts, so it rises as MoE models scale. We introduce Sonar-MoE, which updates each expert tensor during...
Avni Mittal, Avinash Anand, Ashutosh Kumar et al.· 0 citations
Speech-LLMs often exhibit prompt overfitting, where models solely trained on automatic speech recognition (ASR) instruction fail to generalize to new instructions such as speech translation and continue to behave primarily as ASR system. We propose DirectSpeech2LLM, a simple end-to-end framework that preserves the inst...
Hemant Yadav, Sunayana Sitaram, Roger Zimmermann et al.· 0 citations
Nutrition estimation is a fundamental task in consumer diet tracking, clinical dietetics, chronic disease management, sports and hospital nutrition, and broader food computing systems. The existing approaches have progressed along two largely separate axes, vision models that rely on calibrated RGB-depth captures and i...
Aman Kumar, Avinash Anand, Chaitanya Lakhchaura et al.· 0 citations
A five-task diagnostic experiment across physics and geometry benchmarks that isolates failures to perception, reasoning, or both is introduced, and InternS1-mini, despite heavy scientific pretraining and thinking capabilities, falls below the weakest model from experiments on every task.
The work positions the work as a controlled reading of which parts of NCA-style computation carry the weight in language modelling, and examines a hierarchical variant of the Neural Cellular Automaton that cascades three stages with windows.
Avni Mittal, Avinash Anand, Ashutosh Kumar et al.· 0 citations
To solve LLM collaboration with non-language agents, latent state internalization is introduced, which projects the subagent's continuous representations directly into the LLM's token stream as learned state tokens, with dynamic re-encoding as actions advance the environment state.
The proposed InductWave, a wavelet-based inductive embedding method for logical query answering on large KGs, performs on par with the baseline models while having half the number of message-passing layers, and outperforms all of them in most cases.
Mayank Kharbanda, Michael Cochez, R. Shah et al.· arXiv.org· 0 citations