Skip to content

Benchmark

BABILong qa3 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

39.7%
Mean
26%
Median
27.8%
Std dev
17%
10th percentile
90%
90th percentile
13%
Min
13%92%
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-tuneqa3, 64k· BABILong v0 data docs90%
~ Mamba (130M) fine-tuneqa3, 64k· BABILong v0 data docs92%
~ RMT (137M) fine-tuneqa3, 64k· BABILong v0 data docs55%
ChatGLM3 6Bchatglm3-6b-128k, qa3, 64k, 128k· BABILong v0 data docs17%
GPT-4-0125-previewqa3, 64k· BABILong v0 data docs31%
Llama3-ChatQA-1.5-8B + RAGqa3, 64k· BABILong v0 data docs13%
Mixtral 8x22B Instruct v0.1Mixtral-8x22B-Instruct-v0.1, qa3, 64k· BABILong v0 data docs26%
Phi-3 Medium 128K InstructPhi-3-medium-128k-instruct, qa3, 64k· BABILong v0 data docs25%
Phi-3 Mini 128K InstructPhi-3-mini-128k-instruct, qa3, 64k· BABILong v0 data docs22%
Yi-34B01-ai/Yi-34B-200k, qa3, 64k, 200k· BABILong v0 data docs26%

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