Skip to content

Benchmark

BABILong qa1 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

61.7%
Mean
58%
Median
27.7%
Std dev
38%
10th percentile
100%
90th percentile
13%
Min
13%100%
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-tuneqa1, 64k· BABILong v0 data docs100%
~ Mamba (130M) fine-tuneqa1, 64k· BABILong v0 data docs100%
~ RMT (137M) fine-tuneqa1, 64k· BABILong v0 data docs96%
ChatGLM3 6Bchatglm3-6b-128k, qa1, 64k, 128k· BABILong v0 data docs13%
GPT-4-0125-previewqa1, 64k· BABILong v0 data docs58%
Llama3-ChatQA-1.5-8B + RAGqa1, 64k· BABILong v0 data docs56%
Mixtral 8x22B Instruct v0.1Mixtral-8x22B-Instruct-v0.1, qa1, 64k· BABILong v0 data docs38%
Phi-3 Medium 128K InstructPhi-3-medium-128k-instruct, qa1, 64k· BABILong v0 data docs58%
Phi-3 Mini 128K InstructPhi-3-mini-128k-instruct, qa1, 64k· BABILong v0 data docs38%
Yi-34B01-ai/Yi-34B-200k, qa1, 64k, 200k· BABILong v0 data docs60%

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