Salvato in:
| Autore principale: | |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2024
|
| Soggetti: | |
| Accesso online: | https://arxiv.org/abs/2412.04537 |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866913599447891968 |
|---|---|
| author | Bharadwaj, Aryasomayajula Ram |
| author_facet | Bharadwaj, Aryasomayajula Ram |
| contents | Chain-of-Thought (CoT) prompting has significantly enhanced the reasoning abilities of large language models. However, recent studies have shown that models can still perform complex reasoning tasks even when the CoT is replaced with filler(hidden) characters (e.g., "..."), leaving open questions about how models internally process and represent reasoning steps. In this paper, we investigate methods to decode these hidden characters in transformer models trained with filler CoT sequences. By analyzing layer-wise representations using the logit lens method and examining token rankings, we demonstrate that the hidden characters can be recovered without loss of performance. Our findings provide insights into the internal mechanisms of transformer models and open avenues for improving interpretability and transparency in language model reasoning. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2412_04537 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Understanding Hidden Computations in Chain-of-Thought Reasoning Bharadwaj, Aryasomayajula Ram Computation and Language Machine Learning Chain-of-Thought (CoT) prompting has significantly enhanced the reasoning abilities of large language models. However, recent studies have shown that models can still perform complex reasoning tasks even when the CoT is replaced with filler(hidden) characters (e.g., "..."), leaving open questions about how models internally process and represent reasoning steps. In this paper, we investigate methods to decode these hidden characters in transformer models trained with filler CoT sequences. By analyzing layer-wise representations using the logit lens method and examining token rankings, we demonstrate that the hidden characters can be recovered without loss of performance. Our findings provide insights into the internal mechanisms of transformer models and open avenues for improving interpretability and transparency in language model reasoning. |
| title | Understanding Hidden Computations in Chain-of-Thought Reasoning |
| topic | Computation and Language Machine Learning |
| url | https://arxiv.org/abs/2412.04537 |