Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2411.19360 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866917851371143168 |
|---|---|
| author | Dai, Hui Pechi, Dan Yang, Xinyi Banga, Garvit Mantri, Raghav |
| author_facet | Dai, Hui Pechi, Dan Yang, Xinyi Banga, Garvit Mantri, Raghav |
| contents | The Needle-in-a-haystack (NIAH) test is a general task used to assess language models' (LMs') abilities to recall particular information from long input context. This framework however does not provide a means of analyzing what factors, beyond context length, contribute to LMs' abilities or inabilities to separate and recall needles from their haystacks. To provide a systematic means of assessing what features contribute to LMs' NIAH capabilities, we developed a synthetic benchmark called DENIAHL (Data-oriented Evaluation of NIAH for LLM's). Our work expands on previous NIAH studies by ablating NIAH features beyond typical context length including data type, size, and patterns. We find stark differences between GPT-3.5 and LLaMA 2-7B's performance on DENIAHL, and drops in recall performance when features like item size are increased, and to some degree when data type is changed from numbers to letters. This has implications for increasingly large context models, demonstrating factors beyond item-number impact NIAH capabilities. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2411_19360 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | DENIAHL: In-Context Features Influence LLM Needle-In-A-Haystack Abilities Dai, Hui Pechi, Dan Yang, Xinyi Banga, Garvit Mantri, Raghav Computation and Language Artificial Intelligence Machine Learning The Needle-in-a-haystack (NIAH) test is a general task used to assess language models' (LMs') abilities to recall particular information from long input context. This framework however does not provide a means of analyzing what factors, beyond context length, contribute to LMs' abilities or inabilities to separate and recall needles from their haystacks. To provide a systematic means of assessing what features contribute to LMs' NIAH capabilities, we developed a synthetic benchmark called DENIAHL (Data-oriented Evaluation of NIAH for LLM's). Our work expands on previous NIAH studies by ablating NIAH features beyond typical context length including data type, size, and patterns. We find stark differences between GPT-3.5 and LLaMA 2-7B's performance on DENIAHL, and drops in recall performance when features like item size are increased, and to some degree when data type is changed from numbers to letters. This has implications for increasingly large context models, demonstrating factors beyond item-number impact NIAH capabilities. |
| title | DENIAHL: In-Context Features Influence LLM Needle-In-A-Haystack Abilities |
| topic | Computation and Language Artificial Intelligence Machine Learning |
| url | https://arxiv.org/abs/2411.19360 |