Skip to content

Benchmark

BABILong qa2 2k

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

55.6%
Mean
46%
Median
24.6%
Std dev
32%
10th percentile
98.8%
90th percentile
22%
Min
22%99%
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, 2k· BABILong v0 data docs98.8%
~ Mamba (130M) fine-tuneqa2, 2k· BABILong v0 data docs98%
~ RMT (137M) fine-tuneqa2, 2k· BABILong v0 data docs99%
ChatGLM3 6Bchatglm3-6b-128k, qa2, 2k, 128k· BABILong v0 data docs45%
GPT-4-0125-previewqa2, 2k· BABILong v0 data docs72%
Llama 2 7bLLaMA-2-7B-32K, qa2, 2k, 32k· BABILong v0 data docs37%
Llama 3 8B InstructMeta-Llama-3-8B-Instruct, qa2, 2k· BABILong v0 data docs49%
Llama3-ChatQA-1.5-8B + RAGqa2, 2k· BABILong v0 data docs22%
Mistral 7B Instruct v0.2Mistral-7b-Instruct-v0.2, qa2, 2k· BABILong v0 data docs32%
Mixtral 8x22B Instruct v0.1Mixtral-8x22B-Instruct-v0.1, qa2, 2k· BABILong v0 data docs61%
Mixtral 8x7B Instruct v0.1Mixtral-8x7B-Instruct-v0.1, qa2, 2k· BABILong v0 data docs46%
Phi-3 Medium 128K InstructPhi-3-medium-128k-instruct, qa2, 2k· BABILong v0 data docs58%
Phi-3 Mini 128K InstructPhi-3-mini-128k-instruct, qa2, 2k· BABILong v0 data docs38%
Yi-34B01-ai/Yi-34B-200k, qa2, 2k, 200k· BABILong v0 data docs46%
Yi-34B01-ai/Yi-34B, qa2, 2k· BABILong v0 data docs32%

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