Skip to content

Benchmark

BABILong qa1 4k

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

Models and systems measured
14
Scores recorded
15
Sources
1
IQ weight
No weight published

Score distribution

73.9%
Mean
77%
Median
21.8%
Std dev
34%
10th percentile
100%
90th percentile
26%
Min
26%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, 4k· BABILong v0 data docs100%
~ Mamba (130M) fine-tuneqa1, 4k· BABILong v0 data docs100%
~ RMT (137M) fine-tuneqa1, 4k· BABILong v0 data docs100%
ChatGLM3 6Bchatglm3-6b-128k, qa1, 4k, 128k· BABILong v0 data docs72%
GPT-4-0125-previewqa1, 4k· BABILong v0 data docs95%
Llama 2 7bLLaMA-2-7B-32K, qa1, 4k, 32k· BABILong v0 data docs26%
Llama 3 8B InstructMeta-Llama-3-8B-Instruct, qa1, 4k· BABILong v0 data docs79%
Llama3-ChatQA-1.5-8B + RAGqa1, 4k· BABILong v0 data docs58%
Mistral 7B Instruct v0.2Mistral-7b-Instruct-v0.2, qa1, 4k· BABILong v0 data docs64%
Mixtral 8x22B Instruct v0.1Mixtral-8x22B-Instruct-v0.1, qa1, 4k· BABILong v0 data docs89%
Mixtral 8x7B Instruct v0.1Mixtral-8x7B-Instruct-v0.1, qa1, 4k· BABILong v0 data docs77%
Phi-3 Medium 128K InstructPhi-3-medium-128k-instruct, qa1, 4k· BABILong v0 data docs80%
Phi-3 Mini 128K InstructPhi-3-mini-128k-instruct, qa1, 4k· BABILong v0 data docs69%
Yi-34B01-ai/Yi-34B-200k, qa1, 4k, 200k· BABILong v0 data docs66%
Yi-34B01-ai/Yi-34B, qa1, 4k· BABILong v0 data docs34%

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