Skip to content

Benchmark

BABILong qa2 1k

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

58.3%
Mean
51%
Median
23.6%
Std dev
34%
10th percentile
98%
90th percentile
25%
Min
25%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, 1k· BABILong v0 data docs96.9%
~ Mamba (130M) fine-tuneqa2, 1k· BABILong v0 data docs98%
~ RMT (137M) fine-tuneqa2, 1k· BABILong v0 data docs100%
ChatGLM3 6Bchatglm3-6b-128k, qa2, 1k, 128k· BABILong v0 data docs46%
GPT-4-0125-previewqa2, 1k· BABILong v0 data docs79%
Llama 2 7bLLaMA-2-7B-32K, qa2, 1k, 32k· BABILong v0 data docs40%
Llama 3 8B InstructMeta-Llama-3-8B-Instruct, qa2, 1k· BABILong v0 data docs46%
Llama3-ChatQA-1.5-8B + RAGqa2, 1k· BABILong v0 data docs25%
Mistral 7B Instruct v0.2Mistral-7b-Instruct-v0.2, qa2, 1k· BABILong v0 data docs34%
Mixtral 8x22B Instruct v0.1Mixtral-8x22B-Instruct-v0.1, qa2, 1k· BABILong v0 data docs65%
Mixtral 8x7B Instruct v0.1Mixtral-8x7B-Instruct-v0.1, qa2, 1k· BABILong v0 data docs51%
Phi-3 Medium 128K InstructPhi-3-medium-128k-instruct, qa2, 1k· BABILong v0 data docs62%
Phi-3 Mini 128K InstructPhi-3-mini-128k-instruct, qa2, 1k· BABILong v0 data docs38%
Yi-34B01-ai/Yi-34B-200k, qa2, 1k, 200k· BABILong v0 data docs51%
Yi-34B01-ai/Yi-34B, qa2, 1k· BABILong v0 data docs43%

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