Skip to content

Benchmark

BABILong qa1 8k

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

72.1%
Mean
67%
Median
19.3%
Std dev
46%
10th percentile
100%
90th percentile
34%
Min
34%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, 8k· BABILong v0 data docs100%
~ Mamba (130M) fine-tuneqa1, 8k· BABILong v0 data docs100%
~ RMT (137M) fine-tuneqa1, 8k· BABILong v0 data docs100%
ChatGLM3 6Bchatglm3-6b-128k, qa1, 8k, 128k· BABILong v0 data docs67%
GPT-4-0125-previewqa1, 8k· BABILong v0 data docs93%
Llama 2 7bLLaMA-2-7B-32K, qa1, 8k, 32k· BABILong v0 data docs34%
Llama 3 8B InstructMeta-Llama-3-8B-Instruct, qa1, 8k· BABILong v0 data docs62%
Llama3-ChatQA-1.5-8B + RAGqa1, 8k· BABILong v0 data docs58%
Mistral 7B Instruct v0.2Mistral-7b-Instruct-v0.2, qa1, 8k· BABILong v0 data docs63%
Mixtral 8x22B Instruct v0.1Mixtral-8x22B-Instruct-v0.1, qa1, 8k· BABILong v0 data docs79%
Mixtral 8x7B Instruct v0.1Mixtral-8x7B-Instruct-v0.1, qa1, 8k· BABILong v0 data docs65%
Phi-3 Medium 128K InstructPhi-3-medium-128k-instruct, qa1, 8k· BABILong v0 data docs81%
Phi-3 Mini 128K InstructPhi-3-mini-128k-instruct, qa1, 8k· BABILong v0 data docs70%
Yi-34B01-ai/Yi-34B-200k, qa1, 8k, 200k· BABILong v0 data docs63%
Yi-34B01-ai/Yi-34B, qa1, 8k· BABILong v0 data docs46%

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