Skip to content

Benchmark

BABILong qa2 0k

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

66.1%
Mean
62%
Median
22.3%
Std dev
37%
10th percentile
100%
90th percentile
28%
Min
28%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, 0k· BABILong v0 data docs100%
~ Mamba (130M) fine-tuneqa2, 0k· BABILong v0 data docs98%
~ RMT (137M) fine-tuneqa2, 0k· BABILong v0 data docs100%
ChatGLM3 6Bchatglm3-6b-128k, qa2, 0k, 128k· BABILong v0 data docs51%
GPT-4-0125-previewqa2, 0k· BABILong v0 data docs88%
Llama 2 7bLLaMA-2-7B-32K, qa2, 0k, 32k· BABILong v0 data docs37%
Llama 3 8B InstructMeta-Llama-3-8B-Instruct, qa2, 0k· BABILong v0 data docs47%
Llama3-ChatQA-1.5-8B + RAGqa2, 0k· BABILong v0 data docs28%
Mistral 7B Instruct v0.2Mistral-7b-Instruct-v0.2, qa2, 0k· BABILong v0 data docs46%
Mixtral 8x22B Instruct v0.1Mixtral-8x22B-Instruct-v0.1, qa2, 0k· BABILong v0 data docs77%
Mixtral 8x7B Instruct v0.1Mixtral-8x7B-Instruct-v0.1, qa2, 0k· BABILong v0 data docs58%
Phi-3 Medium 128K InstructPhi-3-medium-128k-instruct, qa2, 0k· BABILong v0 data docs76%
Phi-3 Mini 128K InstructPhi-3-mini-128k-instruct, qa2, 0k· BABILong v0 data docs57%
Yi-34B01-ai/Yi-34B-200k, qa2, 0k, 200k· BABILong v0 data docs62%
Yi-34B01-ai/Yi-34B, qa2, 0k· BABILong v0 data docs67%

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