Dataset Viewer
Auto-converted to Parquet Duplicate
model_id
string
dataset_id
string
num_samples
int64
num_rollouts
int64
total_generations
int64
num_moe_layers
int64
num_experts
int64
trained_top_k
int64
max_new_tokens
int64
padding_ignored_in_final_metrics
bool
sampling
dict
seed
int64
dtype
string
total_generation_time_s
float64
accuracy
dict
router_metrics_per_layer
dict
metric_explanations
dict
generated_at
string
deepnevro/LFM2.5-8B-A1B-Expert
deepnevro/MATH-500-Level1
15
1
15
22
32
4
500
true
{ "temperature": 0.2, "top_p": 1, "top_k": 80, "repetition_penalty": 1.05 }
42
bfloat16
305.5
{ "mean": 0.8667, "per_level": { "1": { "count": 15, "accuracy": 0.8667, "mean_output_tokens": 599.27 } }, "per_subject": { "Precalculus": { "count": 2, "accuracy": 1, "mean_output_tokens": 562 }, "Intermediate Algebra": { "count": 2, "accuracy...
{ "2": { "num_real_tokens": 1001, "num_experts": 32, "model_top_k": 4, "logit_mean_mean": -2.106798, "logit_mean_std": 0.245891, "logit_std_mean": 0.86608, "logit_std_std": 0.121633, "activated_k_mean": 4, "activated_k_std": 0, "num_recorded_tokens": 9975, "num_padding_toke...
{ "logit_mean_mean": "Mean of per-token mean across all 32 raw router logits.", "logit_mean_std": "Std of per-token mean logit across real tokens.", "logit_std_mean": "Mean of per-token std across all 32 raw router logits.", "logit_std_std": "Std of per-token logit std across real tokens.", "activated_k_mean"...
2026-07-30T07:16:38.445554Z

No dataset card yet

Downloads last month
34