Skip to content

Benchmark

BABILong qa3 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

31.9%
Mean
24%
Median
27.8%
Std dev
4%
10th percentile
91%
90th percentile
1%
Min
1%95%
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, 32k· BABILong v0 data docs91%
~ Mamba (130M) fine-tuneqa3, 32k· BABILong v0 data docs95%
~ RMT (137M) fine-tuneqa3, 32k· BABILong v0 data docs55%
ChatGLM3 6Bchatglm3-6b-128k, qa3, 32k, 128k· BABILong v0 data docs23%
GPT-4-0125-previewqa3, 32k· BABILong v0 data docs31%
Llama 2 7bLLaMA-2-7B-32K, qa3, 32k· BABILong v0 data docs1%
Llama3-ChatQA-1.5-8B + RAGqa3, 32k· BABILong v0 data docs15%
Mistral 7B Instruct v0.2Mistral-7b-Instruct-v0.2, qa3, 32k· BABILong v0 data docs12%
Mixtral 8x22B Instruct v0.1Mixtral-8x22B-Instruct-v0.1, qa3, 32k· BABILong v0 data docs26%
Mixtral 8x7B Instruct v0.1Mixtral-8x7B-Instruct-v0.1, qa3, 32k· BABILong v0 data docs24%
Phi-3 Medium 128K InstructPhi-3-medium-128k-instruct, qa3, 32k· BABILong v0 data docs27%
Phi-3 Mini 128K InstructPhi-3-mini-128k-instruct, qa3, 32k· BABILong v0 data docs21%
Yi-34B01-ai/Yi-34B-200k, qa3, 32k, 200k· BABILong v0 data docs22%
Yi-34B01-ai/Yi-34B, qa3, 32k· BABILong v0 data docs4%

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