Skip to content

Benchmark

BABILong qa2 32k

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

Models and systems measured
13
Scores recorded
14
Sources
1
IQ weight
No weight published

Score distribution

35.1%
Mean
27%
Median
32.5%
Std dev
4%
10th percentile
98%
90th percentile
2%
Min
2%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-tuneqa2, 32k· BABILong v0 data docs100%
~ Mamba (130M) fine-tuneqa2, 32k· BABILong v0 data docs98%
~ RMT (137M) fine-tuneqa2, 32k· BABILong v0 data docs82%
ChatGLM3 6Bchatglm3-6b-128k, qa2, 32k, 128k· BABILong v0 data docs24%
GPT-4-0125-previewqa2, 32k· BABILong v0 data docs42%
Llama 2 7bLLaMA-2-7B-32K, qa2, 32k· BABILong v0 data docs2%
Llama3-ChatQA-1.5-8B + RAGqa2, 32k· BABILong v0 data docs9%
Mistral 7B Instruct v0.2Mistral-7b-Instruct-v0.2, qa2, 32k· BABILong v0 data docs7%
Mixtral 8x22B Instruct v0.1Mixtral-8x22B-Instruct-v0.1, qa2, 32k· BABILong v0 data docs33%
Mixtral 8x7B Instruct v0.1Mixtral-8x7B-Instruct-v0.1, qa2, 32k· BABILong v0 data docs14%
Phi-3 Medium 128K InstructPhi-3-medium-128k-instruct, qa2, 32k· BABILong v0 data docs27%
Phi-3 Mini 128K InstructPhi-3-mini-128k-instruct, qa2, 32k· BABILong v0 data docs22%
Yi-34B01-ai/Yi-34B-200k, qa2, 32k, 200k· BABILong v0 data docs27%
Yi-34B01-ai/Yi-34B, qa2, 32k· BABILong v0 data docs4%

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