This paper records the process of creating and fine-tuning a conversational large language model (LLM) with the purpose of replicating an individual’s texting style and shares insights on model capacity ceilings, data format design, parameter-efficient fine-tuning, and the cost of manual preference labeling.
Supervised fine-tuning (SFT) equips large language models (LLMs) with specialized capabilities, but often comes at the cost of forgetting the general capabilities of their parent models (i.e., the pretrained models before fine-tuning). This trade-off is especially limiting for queries that require both specialized and...
Ke-Nan Tang, An-Dong Hua, Cheng-Xuan Qian et al.· 0 citations
Looped language models increase effective depth by repeatedly applying a shared block of layers, but existing large-scale recipes require multi-stage training over trillions of tokens, while the benefits of recurrence remain difficult to separate from differences in data and training. In this work, we establish practic...
Andrei Marchenko, В. Б. Безруков, Oleg Kashurin et al.· 4 citations· ⚡1
When a vision-language-action (VLA) model is fine-tuned with its trainable weights in bfloat16 (bf16) and no fp32 master copy, AdamW writes each update straight into bf16, silently rounding away any update smaller than 1/512 of a weight. In 9 of 12 open VLA codebases a default or documented recipe trains this way, 4 on...
Looped large language models trade parameter storage for recurrent computation, but most are trained from scratch. In practice, pretraining with looped depth is expensive, making post-training upcycling of a pretrained non-looped LLM an attractive alternative. We introduce \textbf{UpLoop}, a post-training recipe based...
Tie-Yuan Chen, Hua-Bin Liu, Ying-Tao Zhang et al.· 0 citations
This paper describes participation in all three subtasks, and characterise how much of the enhanced track's lift reflects organiser-rewritten inputs carrying label information rather than genuine argument clarification, and closes with the rare-class limitation on ST3 and the most useful follow-ups.
Stylianos Psychias· Conference and Labs of the E...· 0 citations
Desired-Update-Aligned Synthetic Data (DASA), which uses activation-gradient feedback from a frozen reference model to guide the optimization of continuous synthetic input embeddings, inspired by the role of activation gradients in local risk reduction, targets useful adaptation updates rather than source-text reconstr...
Jin-Hao Zhang, Ze-Yu Liu, Zi-Cheng Yan et al.· arXiv.org· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.