Skip to content

Benchmark

BABILong qa5 64k

Long context benchmarks measure how well models use information in long inputs. Scores are percentages. Higher is better.

Models and systems measured
10
Scores recorded
10
Sources
1
IQ weight
No weight published

Score distribution

76.2%
Mean
78%
Median
14.9%
Std dev
60%
10th percentile
97.9%
90th percentile
55%
Min
55%99%
The band covers the middle 80% of scores. The line marks the median.

Results

The source marks these results as descriptive, not capability scores, so we list them by name instead of ranking them.

Model or systemScore
~ ARMT (137M) fine-tuneqa5, 64k· BABILong v0 data docs97.9%
~ Mamba (130M) fine-tuneqa5, 64k· BABILong v0 data docs99%
~ RMT (137M) fine-tuneqa5, 64k· BABILong v0 data docs89%
ChatGLM3 6Bchatglm3-6b-128k, qa5, 64k, 128k· BABILong v0 data docs55%
GPT-4-0125-previewqa5, 64k· BABILong v0 data docs60%
Llama3-ChatQA-1.5-8B + RAGqa5, 64k· BABILong v0 data docs76%
Mixtral 8x22B Instruct v0.1Mixtral-8x22B-Instruct-v0.1, qa5, 64k· BABILong v0 data docs63%
Phi-3 Medium 128K InstructPhi-3-medium-128k-instruct, qa5, 64k· BABILong v0 data docs78%
Phi-3 Mini 128K InstructPhi-3-mini-128k-instruct, qa5, 64k· BABILong v0 data docs64%
Yi-34B01-ai/Yi-34B-200k, qa5, 64k, 200k· BABILong v0 data docs80%

Data credits: RMT-team, Kuratov et al. Official v0 numeric CSV; results licence not separately stated.