The dataset viewer is not available for this split.
Error code: StreamingRowsError
Exception: CastError
Message: Couldn't cast
stage: string
completed: int64
total: int64
updated_at: string
resolved_revision: string
chat_template_sha256: string
resumed_records: int64
sample_count: int64
weight_dtype: string
model_load_seconds: double
eos_token_id: int64
transformers: string
accelerate: string
started_at: string
model_id: string
finished_at: string
device_map: struct<model.embed_tokens: string, lm_head: string, model.layers.0: string, model.layers.1: string, (... 787 chars omitted)
child 0, model.embed_tokens: string
child 1, lm_head: string
child 2, model.layers.0: string
child 3, model.layers.1: string
child 4, model.layers.2: string
child 5, model.layers.3: string
child 6, model.layers.4: string
child 7, model.layers.5: string
child 8, model.layers.6: string
child 9, model.layers.7: string
child 10, model.layers.8: string
child 11, model.layers.9: string
child 12, model.layers.10: string
child 13, model.layers.11: string
child 14, model.layers.12: string
child 15, model.layers.13: string
child 16, model.layers.14: string
child 17, model.layers.15: string
child 18, model.layers.16: string
child 19, model.layers.17: string
child 20, model.layers.18: string
child 21, model.layers.19: string
child 22, model.layers.20: string
child 23, model.layers.21: string
child 24, model.layers.22: string
child 25, model.layers.23: string
child 26, model.layers.24: string
child 27, model.layers.25: string
child 28, model.layers.26: string
child 29, model.layers.27: string
child 30, model.layers.28: string
child 31, model.layers.29: string
child 32, model.layers.30: string
child 33, model.layers.31: string
child 34, model.norm: string
child 35, model.rotary_emb: string
peak_gpu_allocated_gib: list<item: double>
child 0, item: double
config: struct<model_id: string, revision: string, run_name: string, checkpoint_repo: string, input_sha256: (... 31 chars omitted)
child 0, model_id: string
child 1, revision: string
child 2, run_name: string
child 3, checkpoint_repo: string
child 4, input_sha256: string
child 5, resumed_records: int64
generation_seconds: double
completed_count: int64
python: string
torch: string
quantization: null
gpu_info: list<item: struct<index: int64, name: string, total_gib: double>>
child 0, item: struct<index: int64, name: string, total_gib: double>
child 0, index: int64
child 1, name: string
child 2, total_gib: double
pad_token_id: int64
to
{'config': {'model_id': Value('string'), 'revision': Value('string'), 'run_name': Value('string'), 'checkpoint_repo': Value('string'), 'input_sha256': Value('string'), 'resumed_records': Value('int64')}, 'started_at': Value('string'), 'model_id': Value('string'), 'resolved_revision': Value('string'), 'weight_dtype': Value('string'), 'quantization': Value('null'), 'gpu_info': List({'index': Value('int64'), 'name': Value('string'), 'total_gib': Value('float64')}), 'device_map': {'model.embed_tokens': Value('string'), 'lm_head': Value('string'), 'model.layers.0': Value('string'), 'model.layers.1': Value('string'), 'model.layers.2': Value('string'), 'model.layers.3': Value('string'), 'model.layers.4': Value('string'), 'model.layers.5': Value('string'), 'model.layers.6': Value('string'), 'model.layers.7': Value('string'), 'model.layers.8': Value('string'), 'model.layers.9': Value('string'), 'model.layers.10': Value('string'), 'model.layers.11': Value('string'), 'model.layers.12': Value('string'), 'model.layers.13': Value('string'), 'model.layers.14': Value('string'), 'model.layers.15': Value('string'), 'model.layers.16': Value('string'), 'model.layers.17': Value('string'), 'model.layers.18': Value('string'), 'model.layers.19': Value('string'), 'model.layers.20': Value('string'), 'model.layers.21': Value('string'), 'model.layers.22': Value('string'), 'model.layers.23': Value('string'), 'model.layers.24': Value('string'), 'model.layers.25': Value('string'), 'model.layers.26': Value('string'), 'model.layers.27': Value('string'), 'model.layers.28': Value('string'), 'model.layers.29': Value('string'), 'model.layers.30': Value('string'), 'model.layers.31': Value('string'), 'model.norm': Value('string'), 'model.rotary_emb': Value('string')}, 'python': Value('string'), 'torch': Value('string'), 'transformers': Value('string'), 'accelerate': Value('string'), 'model_load_seconds': Value('float64'), 'chat_template_sha256': Value('string'), 'eos_token_id': Value('int64'), 'pad_token_id': Value('int64'), 'sample_count': Value('int64'), 'resumed_records': Value('int64'), 'finished_at': Value('string'), 'generation_seconds': Value('float64'), 'completed_count': Value('int64'), 'peak_gpu_allocated_gib': List(Value('float64'))}
because column names don't match
Traceback: Traceback (most recent call last):
File "/src/services/worker/src/worker/utils.py", line 149, in get_rows_or_raise
return get_rows(
dataset=dataset,
...<4 lines>...
column_names=column_names,
)
File "/src/libs/libcommon/src/libcommon/utils.py", line 272, in decorator
return func(*args, **kwargs)
File "/src/services/worker/src/worker/utils.py", line 129, in get_rows
rows_plus_one = list(itertools.islice(safe_iter(ds, dataset=dataset), rows_max_number + 1))
File "/src/services/worker/src/worker/utils.py", line 544, in safe_iter
yield from ds.decode(False) if ds.features else ds
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 3052, in __iter__
for key, example in ex_iterable:
^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2559, in __iter__
for key, pa_table in self._iter_arrow():
~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2584, in _iter_arrow
for key, pa_table in self.ex_iterable._iter_arrow():
~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 558, in _iter_arrow
for key, pa_table in iterator:
^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 440, in _iter_arrow
for key, pa_table in self.generate_tables_fn(**gen_kwags):
~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 361, in _generate_tables
self._cast_table(pa_table, json_field_paths=json_field_paths),
~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 132, in _cast_table
pa_table = table_cast(pa_table, self.info.features.arrow_schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2381, in table_cast
return cast_table_to_schema(table, schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2309, in cast_table_to_schema
raise CastError(
...<3 lines>...
)
datasets.table.CastError: Couldn't cast
stage: string
completed: int64
total: int64
updated_at: string
resolved_revision: string
chat_template_sha256: string
resumed_records: int64
sample_count: int64
weight_dtype: string
model_load_seconds: double
eos_token_id: int64
transformers: string
accelerate: string
started_at: string
model_id: string
finished_at: string
device_map: struct<model.embed_tokens: string, lm_head: string, model.layers.0: string, model.layers.1: string, (... 787 chars omitted)
child 0, model.embed_tokens: string
child 1, lm_head: string
child 2, model.layers.0: string
child 3, model.layers.1: string
child 4, model.layers.2: string
child 5, model.layers.3: string
child 6, model.layers.4: string
child 7, model.layers.5: string
child 8, model.layers.6: string
child 9, model.layers.7: string
child 10, model.layers.8: string
child 11, model.layers.9: string
child 12, model.layers.10: string
child 13, model.layers.11: string
child 14, model.layers.12: string
child 15, model.layers.13: string
child 16, model.layers.14: string
child 17, model.layers.15: string
child 18, model.layers.16: string
child 19, model.layers.17: string
child 20, model.layers.18: string
child 21, model.layers.19: string
child 22, model.layers.20: string
child 23, model.layers.21: string
child 24, model.layers.22: string
child 25, model.layers.23: string
child 26, model.layers.24: string
child 27, model.layers.25: string
child 28, model.layers.26: string
child 29, model.layers.27: string
child 30, model.layers.28: string
child 31, model.layers.29: string
child 32, model.layers.30: string
child 33, model.layers.31: string
child 34, model.norm: string
child 35, model.rotary_emb: string
peak_gpu_allocated_gib: list<item: double>
child 0, item: double
config: struct<model_id: string, revision: string, run_name: string, checkpoint_repo: string, input_sha256: (... 31 chars omitted)
child 0, model_id: string
child 1, revision: string
child 2, run_name: string
child 3, checkpoint_repo: string
child 4, input_sha256: string
child 5, resumed_records: int64
generation_seconds: double
completed_count: int64
python: string
torch: string
quantization: null
gpu_info: list<item: struct<index: int64, name: string, total_gib: double>>
child 0, item: struct<index: int64, name: string, total_gib: double>
child 0, index: int64
child 1, name: string
child 2, total_gib: double
pad_token_id: int64
to
{'config': {'model_id': Value('string'), 'revision': Value('string'), 'run_name': Value('string'), 'checkpoint_repo': Value('string'), 'input_sha256': Value('string'), 'resumed_records': Value('int64')}, 'started_at': Value('string'), 'model_id': Value('string'), 'resolved_revision': Value('string'), 'weight_dtype': Value('string'), 'quantization': Value('null'), 'gpu_info': List({'index': Value('int64'), 'name': Value('string'), 'total_gib': Value('float64')}), 'device_map': {'model.embed_tokens': Value('string'), 'lm_head': Value('string'), 'model.layers.0': Value('string'), 'model.layers.1': Value('string'), 'model.layers.2': Value('string'), 'model.layers.3': Value('string'), 'model.layers.4': Value('string'), 'model.layers.5': Value('string'), 'model.layers.6': Value('string'), 'model.layers.7': Value('string'), 'model.layers.8': Value('string'), 'model.layers.9': Value('string'), 'model.layers.10': Value('string'), 'model.layers.11': Value('string'), 'model.layers.12': Value('string'), 'model.layers.13': Value('string'), 'model.layers.14': Value('string'), 'model.layers.15': Value('string'), 'model.layers.16': Value('string'), 'model.layers.17': Value('string'), 'model.layers.18': Value('string'), 'model.layers.19': Value('string'), 'model.layers.20': Value('string'), 'model.layers.21': Value('string'), 'model.layers.22': Value('string'), 'model.layers.23': Value('string'), 'model.layers.24': Value('string'), 'model.layers.25': Value('string'), 'model.layers.26': Value('string'), 'model.layers.27': Value('string'), 'model.layers.28': Value('string'), 'model.layers.29': Value('string'), 'model.layers.30': Value('string'), 'model.layers.31': Value('string'), 'model.norm': Value('string'), 'model.rotary_emb': Value('string')}, 'python': Value('string'), 'torch': Value('string'), 'transformers': Value('string'), 'accelerate': Value('string'), 'model_load_seconds': Value('float64'), 'chat_template_sha256': Value('string'), 'eos_token_id': Value('int64'), 'pad_token_id': Value('int64'), 'sample_count': Value('int64'), 'resumed_records': Value('int64'), 'finished_at': Value('string'), 'generation_seconds': Value('float64'), 'completed_count': Value('int64'), 'peak_gpu_allocated_gib': List(Value('float64'))}
because column names don't matchNeed help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
Aya Hallucination, Factuality, and Selective Abstention: A Rigorous Multi-Benchmark Evaluation & Calibrated Intervention
Lexsi AI Safety & Alignment Hackathon (Track 2: AI Red Teaming, Safety & Alignment)
Models Evaluated: CohereLabs/tiny-aya-global (3.35B) & CohereLabs/aya-expanse-8b (8B)
Author: AstralKS (Hugging Face: ASTRALK | GitHub: AstralKS)
Artifact Collections:
- Hugging Face Model Collection: Aya Safety & Hallucination Evaluations
- Hugging Face Dataset Collection: Aya Safety & Hallucination Benchmarks & Traces
- GitHub Repository: AstralKS/aya-safety-alignment-benchmarks
1. Abstract (<500 words)
Multilingual open-weights foundation models frequently exhibit high surface fluency alongside severe ungrounded hallucination, particularly in knowledge-intensive regimes. In this project for the Lexsi AI Safety & Alignment Hackathon (Track 2: AI Red Teaming, Safety & Alignment), we present a frozen empirical evaluation of Cohere's dense open-weights multilingual modelsβCohereLabs/tiny-aya-global (3.35B) and CohereLabs/aya-expanse-8b (8B)βacross four distinct safety and factuality benchmarks comprising 6,347 evaluated test instances per model (12,694 total model generations). Furthermore, we demonstrate a concrete before-and-after safety alignment intervention (Gate 1) using an integrated pipeline of CuratorKIT, SafeTune, and AuditKIT.
Our baseline battery assesses four complementary axes:
- Short-Form Factuality: SimpleQA Verified (1,000 cases) measuring precision and coverage under factual queries.
- Hallucination vs. False Refusal: HalluLens PreciseWikiQA (1,000 cases) measuring whether models hallucinate ungrounded claims when refusing to abstain.
- Calibrated Selective Abstention: AbstentionBench (978 balanced answerable/unanswerable cases) measuring appropriate abstention behavior and joint task success.
- Metacognitive Self-Awareness: SelfAware (3,369 cases) evaluating whether the model recognizes the boundaries of its own knowledge.
To mitigate ungrounded generation, we implemented a Gate 1 representation-based selective abstention probe extracted from Tiny Aya internal activations (Layer 18), calibrated and routed via SafeTune, and evaluated under AuditKIT. On development cases, Gate 1 increased joint task success from 52.5% to 70.0% (+17.50 percentage points) while retaining 100% of baseline correct answers (14/14) and cutting ungrounded attempts on should-abstain questions by more than half (from 13/20 to 6/20).
All judgments were executed with deterministic greedy decoding using GLM-5.3-Flash as an independent factual judge (6,835 settled raw judgments, zero evaluation exceptions), coupled with exact SimCSE embeddings for semantic unanswerability matching. All datasets, model outputs, judge traces, and reproducible evaluation pipelines are hosted publicly.
2. Before-and-After Intervention: Gate 1 Pilot Results
Using the Hackathon toolkits (CuratorKIT for split hygiene and case extraction, SafeTune for representation-steered abstention routing, and AuditKIT for multi-arm evaluation), we evaluated a prompt-only Layer-18 probe intervention on Tiny Aya Global across balanced answerable/should-abstain cases:
| Development Arm | Joint Task Success β | Correct Retained / 14 β | Attempts / 40 | Should-Abstain Attempts β | False Refusals / 20 β |
|---|---|---|---|---|---|
| Cached Baseline | 21/40 (52.5%) | 14/14 (100%) | 28/40 | 13/20 | 5/20 |
| Gate 1 (Layer-18 Probe + SafeTune) | 28/40 (70.0%) | 14/14 (100%) | 21/40 | 6/20 | 5/20 |
| Question-only TF-IDF Control | 21/40 (52.5%) | 14/14 (100%) | 28/40 | 13/20 | 5/20 |
| Seeded Random Control | 22/40 (55.0%) | 11/14 (78.6%) | 21/40 | 9/20 | 8/20 |
| Always Abstain | 20/40 (50.0%) | 0/14 (0.0%) | 0/40 | 0/20 | 20/20 |
Key Improvements Under Gate 1 Intervention:
- Joint Task Success: Improved from 52.5% β 70.0% (+17.50 percentage points; 95% paired bootstrap interval: [+2.86, +32.61]).
- Behavior Accuracy: Improved from 55.0% β 72.5% (+17.5 percentage points).
- Unknown-label F1: Increased from 43.75% β 71.79% (+28.04 percentage points).
- Zero Loss on Known Facts: Retained 100% of baseline correct answers (14/14).
- Suppression of Ungrounded Guesses: Erroneous attempts on should-abstain questions dropped from 13/20 down to 6/20.
3. Four-Benchmark Baseline Battery
| Benchmark | Target Safety Axis | Primary Metric | Tiny Aya Global | Aya Expanse 8B | N (Rows/model) |
|---|---|---|---|---|---|
| SimpleQA Verified | Short-form factuality | Correct-Answer Yield β | 3.70% | 5.70% | 1,000 |
| Answer Coverage | 95.20% | 95.30% | 1,000 | ||
| Accuracy When Attempted β | 3.89% | 5.98% | 1,000 | ||
| HalluLens PreciseWikiQA | Hallucination vs False Refusal | Hallucination (when not abstained) β | 91.94% | 84.64% | 1,000 |
| False Refusal Rate β | 3.20% | 4.30% | 1,000 | ||
| Correct-Answer Rate β | 7.80% | 14.70% | 1,000 | ||
| AbstentionBench | Calibrated Selective Abstention | Behavior Accuracy β | 68.30% | 72.70% | 978 |
| Joint Task Success β | 51.23% | 60.22% | 978 | ||
| SelfAware | Epistemic Metacognition | Unknown-label F1 β | 34.07% | 43.70% | 3,369 |
| Precision / Recall | 64.67% / 23.08% | 69.79% / 31.81% | 3,369 | ||
| Answerable Reference Accuracy β | 32.04% | 39.01% | 2,338 |
Detailed Denominators & Counts
- SimpleQA Verified (1,000 cases):
- Tiny Aya: 37 correct / 915 incorrect / 48 not attempted.
- Aya Expanse 8B: 57 correct / 896 incorrect / 47 not attempted.
- HalluLens PreciseWikiQA (1,000 cases):
- Tiny Aya: 968 non-abstained / 890 hallucinated / 78 correct / 32 false refusal.
- Aya Expanse 8B: 957 non-abstained / 810 hallucinated / 147 correct / 43 false refusal.
- AbstentionBench (978 balanced cases; 489 answerable, 489 unanswerable):
- Tiny Aya: 668 behavior successes / 501 joint successes.
- Aya Expanse 8B: 711 behavior successes / 589 joint successes.
- SelfAware (3,369 cases; 2,338 answerable, 1,031 unanswerable):
- Tiny Aya: TP 238 / FP 128 / FN 793.
- Aya Expanse 8B: TP 328 / FP 142 / FN 703.
4. Key Findings and Alignment Insights
1. The Factuality Coverage Trap
Both models attempt >95% of factual prompts on SimpleQA, despite having less than a 6% actual correctness rate. In standard generation, Aya models are strongly biased to answer rather than abstain, leading to severe hallucination rates exceeding 84% on PreciseWikiQA.
2. Epistemic Scaling Trends
Scaling from Tiny Aya (3.35B) to Aya Expanse (8B) yields consistent improvements in safety and epistemic calibration:
- Joint task success on AbstentionBench jumps +8.99% (51.23% β 60.22%).
- Unknown-label F1 on SelfAware increases +9.63% (34.07% β 43.70%).
- Hallucination when attempting to answer drops by 7.30% (91.94% β 84.64%). However, even at 8B scale, spontaneous abstention remains inadequate without targeted intervention.
3. Gating Before Generation (Gate 1)
Representation probes extracted from internal model activations (Layer 18) contain separable epistemic signals regarding query familiarity. By acting as an early refusal gate before response autoregression begins, SafeTune effectively suppresses ungrounded generations without degrading known factual recall.
5. Tool Chain Implementation (Lexsi Toolkits)
This work connects three official Lexsi tools in an end-to-end alignment pipeline:
- CuratorKIT: Prepared balanced, group-isolated splits (120 train / 40 calibration / 40 development) preventing question-group leakage across splits.
- SafeTune: Hosted the calibrated Layer-18 probe router in FP32, mapping hidden states to an abstention decision (
I don't know.vs. pass-through). - AuditKIT: Provided standardized evaluation schemas, running 10 multi-arm comparisons across 400 records and 7,600 indicators to measure behavior accuracy and joint task success.
6. Methodology, Provenance, and Reproducibility
- Inference Setup: Greedy one-pass decoding, native chat templates, FP16 precision.
- Judge Setup: GLM-5.3-Flash via hosted API, zero reasoning randomness, strict final-label extraction. Exactly 6,835 evaluation calls completed with zero errors.
- Hash Identifiers:
benchmark-results.jsonSHA-256:c741366272671a0cfd27659ac04a1a64c25dd676a6bad3363068b8c4e90d4955- Tiny Aya responses:
aa7a20232e7652d857433cc93efaf4dcfbc9f212ca28bb3ab6c6775444d9077f - Aya 8B responses:
10cec296f90e9f77be709a7c7643532acf5750786b16047b2e4d9ba1c22175a8
- Cost Accounting: Total actual judge spend is $3.889 across 9,494 settled calls under the strict $18 ceiling.
- Downloads last month
- 28