It is shown that initialization loss and initialization bits-per-byte (Init BPB) are unreliable predictors of downstream convergence, whereas lightweight CPT, as few as 50 steps, provides a cost-effective and reliable signal for selecting the best initialization strategy.
Abstract
Vocabulary extension is an efficient way to adapt pretrained large language models (LLMs) to new languages, but the initialization of newly added token embeddings can strongly affect continued pre-training (CPT) efficiency. We present a systematic study of more than 20 initialization strategies for Hindi vocabulary extension in Nemotron-3-Nano-30B-A3B. Our comparison spans vocabulary-averaging baselines; external and learned initialization methods, including FOCUS, top-k semantic retrieval, and residual MLP mappings; subword composition; norm calibration; and input-output asymmetry. We find that subword composition methods outperform both vocabulary averaging and external/learned initialization approaches. Within subword composition, asymmetric variants achieve the lowest observed early validation loss and reveal distinct preferences for input and output embedding initialization. The best observed configuration initializes the input embedding matrix with uniform subword averaging and Hindi-specific norm calibration, and the output language modeling head with character-length-weighted subword averaging. Relative to the standard Mean-all baseline, this full initialization pipeline reaches comparable validation loss with over a 6x reduction in CPT steps and exceeds the baseline's 3,500-step MILU-Hindi accuracy after only 500 steps. Finally, we show that initialization loss and initialization bits-per-byte (Init BPB) are unreliable predictors of downstream convergence, whereas lightweight CPT, as few as 50 steps, provides a cost-effective and reliable signal for selecting the best initialization strategy.
A trainable input embedding table assigns each vocabulary item an independently adjustable vector. We investigate whether this token-specific parameterization is required for substantial language-modeling capability, or whether a shared Transformer can learn from fixed token identities. We compare three decoder-only la...
An information-weighted cross-entropy loss that rescales token-level contributions using TF-IDF statistics, emphasizing semantically informative tokens while down-weighting ubiquitous ones is presented, offering a lightweight and principled way to mitigate memorization without disrupting standard training dynamics.
Zhi-Jian Li, Stefan Larson, Kevin Leach· 0 citations
Stable 4-bit floating-point (FP4) pretraining is difficult because the E2M1 payload represents only a narrow range of magnitudes. NVIDIA's Transformer Engine \nv{} recipe addresses this with current-tensor scaling, a randomized Hadamard transform (RHT), and bfloat16 (BF16) final layers, adding work outside the FP4 matr...
This work proposes REIGN (Refurbished Embeddings with Integrated Guidance Networks), a contrastively trained bi-encoder that operates on sequences of contextualised chunk embeddings from a frozen Guidance Network rather than on raw tokens for document-to-document retrieval.
Tokenizers are commonly optimized for compression, but a compact vocabulary does not necessarily distribute its capacity evenly across languages. We introduce the Latent Core Tokenizer (LCT), a language-agnostic approach that separates structural discovery from vocabulary construction. LCT uses Minimum Description Leng...
Felermino D. M. A. Ali, Millicent Ochieng, O. Ekwejunor-Etchie et al.· 0 citations
Multilingual Large Language Models (LLMs) traditionally rely on a single vocabulary shared by all supported languages, which can lead to uneven compression across them. Moreover, their large embedding and output matrices increase memory usage and slow inference, notably for small-scale models. It is also wasteful as mo...
Franck Signe Talla, Hippolyte Pilchen, François Yvon et al.· Conference on Empirical Meth...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.