datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
gemma-4-e4b-it-atlas
juiceb0xc0de/gemma-4-e4b-it-atlas
A brain atlas for google/gemma-4-E4B-it, the instruction-tuned E4B member of the Gemma 4 family. This is not a chat dataset or a benchmark. It is an internal-mechanics map built by running activations through a corpus of prompts and scoring what each layer, component, head, and feature direction is doing.
If you want to know what sliding-window and full-attention layers actually do differently inside one model, how KV cache sharing splits a… See the full description on the dataset page: https://huggingface.co/datasets/juiceb0xc0de/gemma-4-e4b-it-atlas.gemma4-e4b-rl100-hf-bf16-sdpa-topk128-overlay
Gemma 4 E4B RL100 top-k-128 target overlay
Precomputed off-policy distillation targets for the E4B-RL-step-100 to E2B experiment.
Source traces: JWei05/gemma4-e4b-rl100-topk128-traces at revision 2b6e49a0a456ee9d67b16a1dc61785562bee90c9
Direction: Gemma 4 E4B RL step 100 teacher to Gemma 4 E2B base student
Target engine: Hugging Face BF16 SDPA full forward
Width: top-k 128
Stored target token IDs: int32
Stored target log-probabilities: float16
Causal alignment: response token… See the full description on the dataset page: https://huggingface.co/datasets/JWei05/gemma4-e4b-rl100-hf-bf16-sdpa-topk128-overlay.gemma-4-e4b-it-atlas-SAE
juiceb0xc0de/gemma-4-e4b-it-atlas-sae
This joins the SAE half of the E4B atlas to the census half. Built from
juiceb0xc0de/gemma-4-e4b-it-SAE-v2
(42 layers, 32x, d_sae 81,920) against
juiceb0xc0de/gemma-4-e4b-it-atlas.
Three splits, all queryable in the browser. No download needed to look around.
The join
The atlas and the SAEs describe the same model in two index spaces that could not talk to each other:
atlas features.feature_idx in [0, 10240) an MLP hidden… See the full description on the dataset page: https://huggingface.co/datasets/juiceb0xc0de/gemma-4-e4b-it-atlas-SAE.TeleQnA-router-gemma4-e4b
TeleQnA router data, Gemma4-E4B
Training data for a quality-estimation classifier over Gemma4-E4B answers on
TeleQnA. Each row is one question on one run: the model's own answer text and a
label saying whether that answer was correct.
Companion to ymoslem/TeleQnA-router,
which holds the same thing for Qwen3-4B-Instruct. The schema is identical, so
the same training script works on either by changing the dataset name.
Split
Rows
Questions
Runs
Accept rate… See the full description on the dataset page: https://huggingface.co/datasets/ymoslem/TeleQnA-router-gemma4-e4b.ConvAI2-ERNIE-enhanced-gemma-4-E4B-it
Visual Memory Results: convai2-ernie-enhanced
This dataset contains the scored output of a visual-memory perplexity experiment.
Experiment metadata
{
"experiment": {
"model_name": "google/gemma-4-E4B-it",
"hf_results_repo": "visual-memory/ConvAI2-ERNIE-enhanced-gemma-4-E4B-it",
"results_jsonl": "results/ConvAI2-ERNIE-enhanced-gemma-4-E4B-it.jsonl",
"hf_dataset": "visual-memory/ConvAI2-With-Ids_1k-no-redundancy",
"hf_mapping_dataset":… See the full description on the dataset page: https://huggingface.co/datasets/visual-memory/ConvAI2-ERNIE-enhanced-gemma-4-E4B-it.ConvAI2-Qwen-original-gemma-4-E4B-it
Visual Memory Results: convai2-qwen-original
This dataset contains the scored output of a visual-memory perplexity experiment.
Experiment metadata
{
"experiment": {
"model_name": "google/gemma-4-E4B-it",
"hf_results_repo": "visual-memory/ConvAI2-Qwen-original-gemma-4-E4B-it",
"results_jsonl": "results/ConvAI2-Qwen-original-gemma-4-E4B-it.jsonl",
"hf_dataset": "visual-memory/ConvAI2-With-Ids_1k-no-redundancy",
"hf_mapping_dataset":… See the full description on the dataset page: https://huggingface.co/datasets/visual-memory/ConvAI2-Qwen-original-gemma-4-E4B-it.ConvAI2-ERNIE-original-gemma-4-E4B-it
Visual Memory Results: convai2-ernie-original
This dataset contains the scored output of a visual-memory perplexity experiment.
Experiment metadata
{
"experiment": {
"model_name": "google/gemma-4-E4B-it",
"hf_results_repo": "visual-memory/ConvAI2-ERNIE-original-gemma-4-E4B-it",
"results_jsonl": "results/ConvAI2-ERNIE-original-gemma-4-E4B-it.jsonl",
"hf_dataset": "visual-memory/ConvAI2-With-Ids_1k-no-redundancy",
"hf_mapping_dataset":… See the full description on the dataset page: https://huggingface.co/datasets/visual-memory/ConvAI2-ERNIE-original-gemma-4-E4B-it.eval_ragtruth-qa_SFT_gemma-4-E4B-it_S130104_epo3_89de_gens_T0_wfs0_s12345_mt512_nosftConvAI2-FLUX-enhanced-gemma-4-E4B-it
Visual Memory Results: convai2-flux-enhanced
This dataset contains the scored output of a visual-memory perplexity experiment.
Experiment metadata
{
"experiment": {
"model_name": "google/gemma-4-E4B-it",
"hf_results_repo": "visual-memory/ConvAI2-FLUX-enhanced-gemma-4-E4B-it",
"results_jsonl": "results/ConvAI2-FLUX-enhanced-gemma-4-E4B-it.jsonl",
"hf_dataset": "visual-memory/ConvAI2-With-Ids_1k-no-redundancy",
"hf_mapping_dataset":… See the full description on the dataset page: https://huggingface.co/datasets/visual-memory/ConvAI2-FLUX-enhanced-gemma-4-E4B-it.PersonaChat-FLUX-enhanced-gemma-4-E4B-it
Visual Memory Results: personachat-flux-enhanced
This dataset contains the scored output of a visual-memory perplexity experiment.
Experiment metadata
{
"experiment": {
"model_name": "google/gemma-4-E4B-it",
"hf_results_repo": "visual-memory/PersonaChat-FLUX-enhanced-gemma-4-E4B-it",
"results_jsonl": "results/PersonaChat-FLUX-enhanced-gemma-4-E4B-it.jsonl",
"hf_dataset": "visual-memory/PersonaChat-With-Ids_1k-no-redundancy"… See the full description on the dataset page: https://huggingface.co/datasets/visual-memory/PersonaChat-FLUX-enhanced-gemma-4-E4B-it.ConvAI2-Qwen-enhanced-gemma-4-E4B-it
Visual Memory Results: convai2-qwen-enhanced
This dataset contains the scored output of a visual-memory perplexity experiment.
Experiment metadata
{
"experiment": {
"model_name": "google/gemma-4-E4B-it",
"hf_results_repo": "visual-memory/ConvAI2-Qwen-enhanced-gemma-4-E4B-it",
"results_jsonl": "results/ConvAI2-Qwen-enhanced-gemma-4-E4B-it.jsonl",
"hf_dataset": "visual-memory/ConvAI2-With-Ids_1k-no-redundancy",
"hf_mapping_dataset":… See the full description on the dataset page: https://huggingface.co/datasets/visual-memory/ConvAI2-Qwen-enhanced-gemma-4-E4B-it.Synthetic-Persona-Chat-ERNIE-enhanced-gemma-4-E4B-it
Visual Memory Results: synthetic-persona-chat-ernie-enhanced
This dataset contains the scored output of a visual-memory perplexity experiment.
Experiment metadata
{
"experiment": {
"model_name": "google/gemma-4-E4B-it",
"hf_results_repo": "visual-memory/Synthetic-Persona-Chat-ERNIE-enhanced-gemma-4-E4B-it",
"results_jsonl": "results/Synthetic-Persona-Chat-ERNIE-enhanced-gemma-4-E4B-it.jsonl",
"hf_dataset":… See the full description on the dataset page: https://huggingface.co/datasets/visual-memory/Synthetic-Persona-Chat-ERNIE-enhanced-gemma-4-E4B-it.PersonaChat-Qwen-original-gemma-4-E4B-it
Visual Memory Results: personachat-qwen-original
This dataset contains the scored output of a visual-memory perplexity experiment.
Experiment metadata
{
"experiment": {
"model_name": "google/gemma-4-E4B-it",
"hf_results_repo": "visual-memory/PersonaChat-Qwen-original-gemma-4-E4B-it",
"results_jsonl": "results/PersonaChat-Qwen-original-gemma-4-E4B-it.jsonl",
"hf_dataset": "visual-memory/PersonaChat-With-Ids_1k-no-redundancy"… See the full description on the dataset page: https://huggingface.co/datasets/visual-memory/PersonaChat-Qwen-original-gemma-4-E4B-it.PersonaChat-ERNIE-original-gemma-4-E4B-it
Visual Memory Results: personachat-ernie-original
This dataset contains the scored output of a visual-memory perplexity experiment.
Experiment metadata
{
"experiment": {
"model_name": "google/gemma-4-E4B-it",
"hf_results_repo": "visual-memory/PersonaChat-ERNIE-original-gemma-4-E4B-it",
"results_jsonl": "results/PersonaChat-ERNIE-original-gemma-4-E4B-it.jsonl",
"hf_dataset": "visual-memory/PersonaChat-With-Ids_1k-no-redundancy"… See the full description on the dataset page: https://huggingface.co/datasets/visual-memory/PersonaChat-ERNIE-original-gemma-4-E4B-it.ConvAI2-FLUX-original-gemma-4-E4B-it
Visual Memory Results: convai2-flux-original
This dataset contains the scored output of a visual-memory perplexity experiment.
Experiment metadata
{
"experiment": {
"model_name": "google/gemma-4-E4B-it",
"hf_results_repo": "visual-memory/ConvAI2-FLUX-original-gemma-4-E4B-it",
"results_jsonl": "results/ConvAI2-FLUX-original-gemma-4-E4B-it.jsonl",
"hf_dataset": "visual-memory/ConvAI2-With-Ids_1k-no-redundancy",
"hf_mapping_dataset":… See the full description on the dataset page: https://huggingface.co/datasets/visual-memory/ConvAI2-FLUX-original-gemma-4-E4B-it.Synthetic-Persona-Chat-FLUX-original-gemma-4-E4B-it
Visual Memory Results: synthetic-persona-chat-flux-original
This dataset contains the scored output of a visual-memory perplexity experiment.
Experiment metadata
{
"experiment": {
"model_name": "google/gemma-4-E4B-it",
"hf_results_repo": "visual-memory/Synthetic-Persona-Chat-FLUX-original-gemma-4-E4B-it",
"results_jsonl": "results/Synthetic-Persona-Chat-FLUX-original-gemma-4-E4B-it.jsonl",
"hf_dataset":… See the full description on the dataset page: https://huggingface.co/datasets/visual-memory/Synthetic-Persona-Chat-FLUX-original-gemma-4-E4B-it.Synthetic-Persona-Chat-FLUX-enhanced-gemma-4-E4B-it
Visual Memory Results: synthetic-persona-chat-flux-enhanced
This dataset contains the scored output of a visual-memory perplexity experiment.
Experiment metadata
{
"experiment": {
"model_name": "google/gemma-4-E4B-it",
"hf_results_repo": "visual-memory/Synthetic-Persona-Chat-FLUX-enhanced-gemma-4-E4B-it",
"results_jsonl": "results/Synthetic-Persona-Chat-FLUX-enhanced-gemma-4-E4B-it.jsonl",
"hf_dataset":… See the full description on the dataset page: https://huggingface.co/datasets/visual-memory/Synthetic-Persona-Chat-FLUX-enhanced-gemma-4-E4B-it.Synthetic-Persona-Chat-ERNIE-original-gemma-4-E4B-it
Visual Memory Results: synthetic-persona-chat-ernie-original
This dataset contains the scored output of a visual-memory perplexity experiment.
Experiment metadata
{
"experiment": {
"model_name": "google/gemma-4-E4B-it",
"hf_results_repo": "visual-memory/Synthetic-Persona-Chat-ERNIE-original-gemma-4-E4B-it",
"results_jsonl": "results/Synthetic-Persona-Chat-ERNIE-original-gemma-4-E4B-it.jsonl",
"hf_dataset":… See the full description on the dataset page: https://huggingface.co/datasets/visual-memory/Synthetic-Persona-Chat-ERNIE-original-gemma-4-E4B-it.PersonaChat-Qwen-enhanced-gemma-4-E4B-it
Visual Memory Results: personachat-qwen-enhanced
This dataset contains the scored output of a visual-memory perplexity experiment.
Experiment metadata
{
"experiment": {
"model_name": "google/gemma-4-E4B-it",
"hf_results_repo": "visual-memory/PersonaChat-Qwen-enhanced-gemma-4-E4B-it",
"results_jsonl": "results/PersonaChat-Qwen-enhanced-gemma-4-E4B-it.jsonl",
"hf_dataset": "visual-memory/PersonaChat-With-Ids_1k-no-redundancy"… See the full description on the dataset page: https://huggingface.co/datasets/visual-memory/PersonaChat-Qwen-enhanced-gemma-4-E4B-it.Synthetic-Persona-Chat-Qwen-original-gemma-4-E4B-it
Visual Memory Results: synthetic-persona-chat-qwen-original
This dataset contains the scored output of a visual-memory perplexity experiment.
Experiment metadata
{
"experiment": {
"model_name": "google/gemma-4-E4B-it",
"hf_results_repo": "visual-memory/Synthetic-Persona-Chat-Qwen-original-gemma-4-E4B-it",
"results_jsonl": "results/Synthetic-Persona-Chat-Qwen-original-gemma-4-E4B-it.jsonl",
"hf_dataset":… See the full description on the dataset page: https://huggingface.co/datasets/visual-memory/Synthetic-Persona-Chat-Qwen-original-gemma-4-E4B-it.eval_ragtruth-qa_PERL_gemma-4-E4B-it_S130104_ace3_gens_T0_wfs0_s12345_mt512_sft1c9bb9PersonaChat-ERNIE-enhanced-gemma-4-E4B-it
Visual Memory Results: personachat-ernie-enhanced
This dataset contains the scored output of a visual-memory perplexity experiment.
Experiment metadata
{
"experiment": {
"model_name": "google/gemma-4-E4B-it",
"hf_results_repo": "visual-memory/PersonaChat-ERNIE-enhanced-gemma-4-E4B-it",
"results_jsonl": "results/PersonaChat-ERNIE-enhanced-gemma-4-E4B-it.jsonl",
"hf_dataset": "visual-memory/PersonaChat-With-Ids_1k-no-redundancy"… See the full description on the dataset page: https://huggingface.co/datasets/visual-memory/PersonaChat-ERNIE-enhanced-gemma-4-E4B-it.Synthetic-Persona-Chat-Qwen-enhanced-gemma-4-E4B-it
Visual Memory Results: synthetic-persona-chat-qwen-enhanced
This dataset contains the scored output of a visual-memory perplexity experiment.
Experiment metadata
{
"experiment": {
"model_name": "google/gemma-4-E4B-it",
"hf_results_repo": "visual-memory/Synthetic-Persona-Chat-Qwen-enhanced-gemma-4-E4B-it",
"results_jsonl": "results/Synthetic-Persona-Chat-Qwen-enhanced-gemma-4-E4B-it.jsonl",
"hf_dataset":… See the full description on the dataset page: https://huggingface.co/datasets/visual-memory/Synthetic-Persona-Chat-Qwen-enhanced-gemma-4-E4B-it.rebase_thinkprm_gemma-4-E4B-it_lcb_v6_ns32_md4_bt0_1_seed42_lcb_v6_thinkprm_2gpu
Aggregated Metrics
Weighted mean of per-shard wandb summary values (weights = shard rows). See _meta/aggregated_metrics.json for raw values and _meta/splits.json for the per-shard wandb run names.
Aggregated from 16 shards.
Metric
Value
avg_response_tokens
7926.66
generation_phase_time_s
2617.85
judge_output_tokens_level_1
2.88528e+06
judge_output_tokens_level_2
879875
judge_output_tokens_level_3
159472
judge_output_tokens_level_4
27007.8
maj@1
0
maj@16
0… See the full description on the dataset page: https://huggingface.co/datasets/anirudhb11/rebase_thinkprm_gemma-4-E4B-it_lcb_v6_ns32_md4_bt0_1_seed42_lcb_v6_thinkprm_2gpu.eval_gemma-4-E4B-it_gens_T0_wfs0PersonaChat-FLUX-original-gemma-4-E4B-it
Visual Memory Results: personachat-flux-original
This dataset contains the scored output of a visual-memory perplexity experiment.
Experiment metadata
{
"experiment": {
"model_name": "google/gemma-4-E4B-it",
"hf_results_repo": "visual-memory/PersonaChat-FLUX-original-gemma-4-E4B-it",
"results_jsonl": "results/PersonaChat-FLUX-original-gemma-4-E4B-it.jsonl",
"hf_dataset": "visual-memory/PersonaChat-With-Ids_1k-no-redundancy"… See the full description on the dataset page: https://huggingface.co/datasets/visual-memory/PersonaChat-FLUX-original-gemma-4-E4B-it.eh-gemma4-e4b-kv-seam-quarantine
gemma4-e4b-kv-seam-quarantine -- aggregate exhaust
Aggregate-only: every file committed under this experiment's analysis-committed/ tree (dose-response tables, direction fits, gate AUROCs, manifests, and any other analysis artifact), copied byte-for-byte. No source question text, aliases, or per-row generation text -- analysis-committed/ never carries those.
HF repo: professorsynapse/eh-gemma4-e4b-kv-seam-quarantine
Provenance
Experiment:… See the full description on the dataset page: https://huggingface.co/datasets/professorsynapse/eh-gemma4-e4b-kv-seam-quarantine.eval_bosch_gemma-4-E4B-it_gens_T1_wfs2_s12345_mt512_nosfteval_npov_gemma-4-E4B-it_gens_T0_3_wfs0_s12345_mt256_nosfteval_bosch_gemma-4-E4B-it_gens_T1_wfs0_s12345_mt512_nosft