The dataset viewer is not available for this split.
Error code: StreamingRowsError
Exception: CastError
Message: Couldn't cast
</think>: int64
</tool_call>: int64
</tool_response>: int64
<think>: int64
<tool_call>: int64
<tool_response>: int64
<|box_end|>: int64
<|box_start|>: int64
<|endoftext|>: int64
<|file_sep|>: int64
<|fim_middle|>: int64
<|fim_pad|>: int64
<|fim_prefix|>: int64
<|fim_suffix|>: int64
<|im_end|>: int64
<|im_start|>: int64
<|image_pad|>: int64
<|object_ref_end|>: int64
<|object_ref_start|>: int64
<|quad_end|>: int64
<|quad_start|>: int64
<|repo_name|>: int64
<|video_pad|>: int64
<|vision_end|>: int64
<|vision_pad|>: int64
<|vision_start|>: int64
revision: null
fan_in_fan_out: bool
loftq_config: struct<>
corda_config: null
use_dora: bool
task_type: string
base_model_name_or_path: string
inference_mode: bool
target_modules: list<item: string>
child 0, item: string
layers_pattern: null
alora_invocation_tokens: null
alpha_pattern: struct<>
auto_mapping: null
eva_config: null
megatron_core: string
exclude_modules: null
bias: string
target_parameters: null
peft_type: string
arrow_config: null
modules_to_save: null
layers_to_transform: null
lora_ga_config: null
use_bdlora: null
layer_replication: null
lora_alpha: int64
lora_dropout: double
qalora_group_size: int64
rank_pattern: struct<>
peft_version: string
use_qalora: bool
ensure_weight_tying: bool
init_lora_weights: bool
lora_bias: bool
r: int64
trainable_token_indices: null
megatron_config: null
use_rslora: bool
to
{'alora_invocation_tokens': Value('null'), 'alpha_pattern': {}, 'arrow_config': Value('null'), 'auto_mapping': Value('null'), 'base_model_name_or_path': Value('string'), 'bias': Value('string'), 'corda_config': Value('null'), 'ensure_weight_tying': Value('bool'), 'eva_config': Value('null'), 'exclude_modules': Value('null'), 'fan_in_fan_out': Value('bool'), 'inference_mode': Value('bool'), 'init_lora_weights': Value('bool'), 'layer_replication': Value('null'), 'layers_pattern': Value('null'), 'layers_to_transform': Value('null'), 'loftq_config': {}, 'lora_alpha': Value('int64'), 'lora_bias': Value('bool'), 'lora_dropout': Value('float64'), 'lora_ga_config': Value('null'), 'megatron_config': Value('null'), 'megatron_core': Value('string'), 'modules_to_save': Value('null'), 'peft_type': Value('string'), 'peft_version': Value('string'), 'qalora_group_size': Value('int64'), 'r': Value('int64'), 'rank_pattern': {}, 'revision': Value('null'), 'target_modules': List(Value('string')), 'target_parameters': Value('null'), 'task_type': Value('string'), 'trainable_token_indices': Value('null'), 'use_bdlora': Value('null'), 'use_dora': Value('bool'), 'use_qalora': Value('bool'), 'use_rslora': Value('bool')}
because column names don't match
Traceback: Traceback (most recent call last):
File "/src/services/worker/src/worker/utils.py", line 147, 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 127, 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 483, 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 2840, in __iter__
for key, example in ex_iterable:
^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2373, in __iter__
for key, pa_table in self._iter_arrow():
~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2398, 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 536, in _iter_arrow
for key, pa_table in iterator:
^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 419, 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 343, 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 2378, in table_cast
return cast_table_to_schema(table, schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2306, in cast_table_to_schema
raise CastError(
...<3 lines>...
)
datasets.table.CastError: Couldn't cast
</think>: int64
</tool_call>: int64
</tool_response>: int64
<think>: int64
<tool_call>: int64
<tool_response>: int64
<|box_end|>: int64
<|box_start|>: int64
<|endoftext|>: int64
<|file_sep|>: int64
<|fim_middle|>: int64
<|fim_pad|>: int64
<|fim_prefix|>: int64
<|fim_suffix|>: int64
<|im_end|>: int64
<|im_start|>: int64
<|image_pad|>: int64
<|object_ref_end|>: int64
<|object_ref_start|>: int64
<|quad_end|>: int64
<|quad_start|>: int64
<|repo_name|>: int64
<|video_pad|>: int64
<|vision_end|>: int64
<|vision_pad|>: int64
<|vision_start|>: int64
revision: null
fan_in_fan_out: bool
loftq_config: struct<>
corda_config: null
use_dora: bool
task_type: string
base_model_name_or_path: string
inference_mode: bool
target_modules: list<item: string>
child 0, item: string
layers_pattern: null
alora_invocation_tokens: null
alpha_pattern: struct<>
auto_mapping: null
eva_config: null
megatron_core: string
exclude_modules: null
bias: string
target_parameters: null
peft_type: string
arrow_config: null
modules_to_save: null
layers_to_transform: null
lora_ga_config: null
use_bdlora: null
layer_replication: null
lora_alpha: int64
lora_dropout: double
qalora_group_size: int64
rank_pattern: struct<>
peft_version: string
use_qalora: bool
ensure_weight_tying: bool
init_lora_weights: bool
lora_bias: bool
r: int64
trainable_token_indices: null
megatron_config: null
use_rslora: bool
to
{'alora_invocation_tokens': Value('null'), 'alpha_pattern': {}, 'arrow_config': Value('null'), 'auto_mapping': Value('null'), 'base_model_name_or_path': Value('string'), 'bias': Value('string'), 'corda_config': Value('null'), 'ensure_weight_tying': Value('bool'), 'eva_config': Value('null'), 'exclude_modules': Value('null'), 'fan_in_fan_out': Value('bool'), 'inference_mode': Value('bool'), 'init_lora_weights': Value('bool'), 'layer_replication': Value('null'), 'layers_pattern': Value('null'), 'layers_to_transform': Value('null'), 'loftq_config': {}, 'lora_alpha': Value('int64'), 'lora_bias': Value('bool'), 'lora_dropout': Value('float64'), 'lora_ga_config': Value('null'), 'megatron_config': Value('null'), 'megatron_core': Value('string'), 'modules_to_save': Value('null'), 'peft_type': Value('string'), 'peft_version': Value('string'), 'qalora_group_size': Value('int64'), 'r': Value('int64'), 'rank_pattern': {}, 'revision': Value('null'), 'target_modules': List(Value('string')), 'target_parameters': Value('null'), 'task_type': Value('string'), 'trainable_token_indices': Value('null'), 'use_bdlora': Value('null'), 'use_dora': Value('bool'), 'use_qalora': Value('bool'), 'use_rslora': Value('bool')}
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.
2026.RA.Branch-Backtrack-GRPO — rewinding no-deal negotiations as GRPO exploration
Campaign artifacts for branch-backtracking GRPO (Qwen3-8B LoRA negotiation policy): mine decision nodes
from saved no-deal episodes, replay the exact prefix (interlens.arena.replay.apply_prefix), sample K=8
on-policy continuations per node, and train on within-node advantages. Proposal, code, and the results note
(0063) live in the source repo; the preregistered POSITIVE gate failed — at ~10% branch gradient mass the
arm collapses held-out closure at rung 15 like the no-branch λ=0 control — while the viability result stands:
40% of late-game no-deal states are rescuable by the policy's own resampling (bimodal: dead ends or
near-misses), and branch waves reliably produce signed advantages.
Headline (held-out primary bank, paired vs a hardware-matched base, deal-rate Δ, computed from
analysis/verdict_ckpt15.json at package build time):
{ "branch25_ckpt5": { "estimate": -0.0763888888888889, "ci_low": -0.13194444444444445, "ci_high": -0.024218750000000473, "n_pairs": 288, "n_clusters": 48, "pairing": "instance+seed+arm" }, "lam0m_ckpt5": { "estimate": -0.008333333333333333, "ci_low": -0.04375, "ci_high": 0.025, "n_pairs": 480, "n_clusters": 48, "pairing": "instance+seed+arm" }, "branch25_ckpt15": { "estimate": -0.3125, "ci_low": -0.40625, "ci_high": -0.22916666666666666, "n_pairs": 96, "n_clusters": 48, "pairing": "instance+seed+arm" }, "lam0m_ckpt15": { "estimate": -0.265625, "ci_low": -0.3489583333333333, "ci_high": -0.1875, "n_pairs": 192, "n_clusters": 48, "pairing": "instance+seed+arm" } }
Contents: pool/ (554 replay-verified branch nodes), pilot/ (160 base-policy continuations),
training/ (both run lineages: telemetry, wave transcripts, LoRA checkpoints 5/10/15/20/25),
eval/ (one jsonl.gz per complete eval shard; partial shards excluded for censoring),
analysis/ (Phase-0 gate, preliminary and final contrasts). Episode schema: interlens Episode.to_json().
Build counts: {
"pilot_episodes": 160,
"branch25/step0005": 96,
"branch25/step0010": 24,
"branch25_resume10/step0015": 80,
"branch25_resume10/step0020": 24,
"branch25_resume10/step0025": 80,
"eval/baseline_hw_primary_s0": 96,
"eval/baseline_hw_primary_s1": 96,
"eval/baseline_hw_primary_s2": 96,
"eval/baseline_hw_primary_s3": 96,
"eval/baseline_hw_primary_s4": 96,
"eval/lam0m_ckpt15_primary_s0": 96,
"eval/lam0m_ckpt15_primary_s1": 96,
"eval/lam0m_ckpt15_primary_s2": -48,
"eval/lam0m_ckpt5_primary_s0": 96,
"eval/lam0m_ckpt5_primary_s1": 96,
"eval/lam0m_ckpt5_primary_s2": 96,
"eval/lam0m_ckpt5_primary_s3": 96,
"eval/lam0m_ckpt5_primary_s4": 96,
"eval/branch25_ckpt15_primary_s0": 96,
"eval/branch25_ckpt15_primary_s1": -1,
"eval/branch25_ckpt25_primary_s0": -10,
"eval/branch25_ckpt5_primary_s0": 96,
"eval/branch25_ckpt5_primary_s1": 96,
"eval/branch25_ckpt5_primary_s2": 96,
"eval/branch25_ckpt5_primary_s3": -37
}
- Downloads last month
- 56