Skip to content

Benchmark

BABILong qa3 16k

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

37.4%
Mean
27%
Median
26.5%
Std dev
17%
10th percentile
92%
90th percentile
12%
Min
12%96%
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, 16k· BABILong v0 data docs92%
~ Mamba (130M) fine-tuneqa3, 16k· BABILong v0 data docs96%
~ RMT (137M) fine-tuneqa3, 16k· BABILong v0 data docs66%
ChatGLM3 6Bchatglm3-6b-128k, qa3, 16k, 128k· BABILong v0 data docs25%
GPT-4-0125-previewqa3, 16k· BABILong v0 data docs45%
Llama 2 7bLLaMA-2-7B-32K, qa3, 16k, 32k· BABILong v0 data docs21%
Llama3-ChatQA-1.5-8B + RAGqa3, 16k· BABILong v0 data docs17%
Mistral 7B Instruct v0.2Mistral-7b-Instruct-v0.2, qa3, 16k· BABILong v0 data docs18%
Mixtral 8x22B Instruct v0.1Mixtral-8x22B-Instruct-v0.1, qa3, 16k· BABILong v0 data docs27%
Mixtral 8x7B Instruct v0.1Mixtral-8x7B-Instruct-v0.1, qa3, 16k· BABILong v0 data docs29%
Phi-3 Medium 128K InstructPhi-3-medium-128k-instruct, qa3, 16k· BABILong v0 data docs30%
Phi-3 Mini 128K InstructPhi-3-mini-128k-instruct, qa3, 16k· BABILong v0 data docs24%
Yi-34B01-ai/Yi-34B-200k, qa3, 16k, 200k· BABILong v0 data docs22%
Yi-34B01-ai/Yi-34B, qa3, 16k· BABILong v0 data docs12%

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