Skip to content

Benchmark

BABILong qa4 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

63.9%
Mean
59%
Median
17%
Std dev
50%
10th percentile
99.8%
90th percentile
42%
Min
42%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-tuneqa4, 2k· BABILong v0 data docs99.8%
~ Mamba (130M) fine-tuneqa4, 2k· BABILong v0 data docs100%
~ RMT (137M) fine-tuneqa4, 2k· BABILong v0 data docs87%
ChatGLM3 6Bchatglm3-6b-128k, qa4, 2k, 128k· BABILong v0 data docs42%
GPT-4-0125-previewqa4, 2k· BABILong v0 data docs63%
Llama 2 7bLLaMA-2-7B-32K, qa4, 2k, 32k· BABILong v0 data docs51%
Llama 3 8B InstructMeta-Llama-3-8B-Instruct, qa4, 2k· BABILong v0 data docs50%
Llama3-ChatQA-1.5-8B + RAGqa4, 2k· BABILong v0 data docs56%
Mistral 7B Instruct v0.2Mistral-7b-Instruct-v0.2, qa4, 2k· BABILong v0 data docs58%
Mixtral 8x22B Instruct v0.1Mixtral-8x22B-Instruct-v0.1, qa4, 2k· BABILong v0 data docs59%
Mixtral 8x7B Instruct v0.1Mixtral-8x7B-Instruct-v0.1, qa4, 2k· BABILong v0 data docs59%
Phi-3 Medium 128K InstructPhi-3-medium-128k-instruct, qa4, 2k· BABILong v0 data docs63%
Phi-3 Mini 128K InstructPhi-3-mini-128k-instruct, qa4, 2k· BABILong v0 data docs56%
Yi-34B01-ai/Yi-34B-200k, qa4, 2k, 200k· BABILong v0 data docs64%
Yi-34B01-ai/Yi-34B, qa4, 2k· BABILong v0 data docs51%

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