Skip to content

Benchmark

BABILong qa1 32k

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

Models and systems measured
13
Scores recorded
14
Sources
1
IQ weight
No weight published

Score distribution

56.6%
Mean
60%
Median
29.3%
Std dev
4%
10th percentile
100%
90th percentile
3%
Min
3%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, 32k· BABILong v0 data docs100%
~ Mamba (130M) fine-tuneqa1, 32k· BABILong v0 data docs100%
~ RMT (137M) fine-tuneqa1, 32k· BABILong v0 data docs99%
ChatGLM3 6Bchatglm3-6b-128k, qa1, 32k, 128k· BABILong v0 data docs47%
GPT-4-0125-previewqa1, 32k· BABILong v0 data docs62%
Llama 2 7bLLaMA-2-7B-32K, qa1, 32k· BABILong v0 data docs3%
Llama3-ChatQA-1.5-8B + RAGqa1, 32k· BABILong v0 data docs60%
Mistral 7B Instruct v0.2Mistral-7b-Instruct-v0.2, qa1, 32k· BABILong v0 data docs45%
Mixtral 8x22B Instruct v0.1Mixtral-8x22B-Instruct-v0.1, qa1, 32k· BABILong v0 data docs40%
Mixtral 8x7B Instruct v0.1Mixtral-8x7B-Instruct-v0.1, qa1, 32k· BABILong v0 data docs49%
Phi-3 Medium 128K InstructPhi-3-medium-128k-instruct, qa1, 32k· BABILong v0 data docs69%
Phi-3 Mini 128K InstructPhi-3-mini-128k-instruct, qa1, 32k· BABILong v0 data docs53%
Yi-34B01-ai/Yi-34B-200k, qa1, 32k, 200k· BABILong v0 data docs62%
Yi-34B01-ai/Yi-34B, qa1, 32k· BABILong v0 data docs4%

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