Failed episodes exhibit systematically more negative temporal curvature kappa across all six task-LLM pairings, with correct minus incorrect mean differences of: Math (Llama-3.1-8B) +0.047, Code (Qwen3-14B) +0.148, Retail (Qwen3-14B) +0.030, Airline (Qwen3-14B) +0.049, Retail (Qwen3-32B) +0.035, and Airline (Qwen3-32B) +0.088.整组结论：实验受阻，详见具体原因0/6 项指标获得支持 · 0 项已有评估6项测量结果展开结果表

The offline episode traces, prompt sharding scripts for Math and Code from Lost in Conversation, and exact hidden state trajectory extraction codebase are not provided in the paper snapshot or any public repository.

测量结果

已评估 0/6 项

获得支持 0有争议或冲突 0尚未评估 6

保留同一结论下的完整结果组。已有评估排在前面；单项判断不替代整组结论。展开行查看条件、来源与历次证据。

每行是一项测量，分别显示论文报告、最近可用实测和累计证据状态。
指标与条件	论文报告	最近可用实测	证据判断
task: Math · model: Llama-3.1-8B	0.047 scalar	—	尚未评估
task: Code · model: Qwen3-14B	0.148 scalar	—	尚未评估
task: Retail · model: Qwen3-14B	0.03 scalar	—	尚未评估
task: Airline · model: Qwen3-14B	0.049 scalar	—	尚未评估
task: Retail · model: Qwen3-32B	0.035 scalar	—	尚未评估
task: Airline · model: Qwen3-32B	0.088 scalar	—	尚未评估

打开 Claim 详情

The learned classifier predicting the agent's next action achieves F1 scores of 0.826 for Read, 0.711 for Respond, 0.700 for Write, and 0.531 for Transfer.整组结论：实验受阻，详见具体原因0/4 项指标获得支持 · 0 项已有评估4项测量结果展开结果表

The paper does not provide the classifier model, training code, loss function, evaluation split, or feature representation.

测量结果

已评估 0/4 项

获得支持 0有争议或冲突 0尚未评估 4

保留同一结论下的完整结果组。已有评估排在前面；单项判断不替代整组结论。展开行查看条件、来源与历次证据。

每行是一项测量，分别显示论文报告、最近可用实测和累计证据状态。
指标与条件	论文报告	最近可用实测	证据判断
action: Read	0.826 fraction	—	尚未评估
action: Respond	0.711 fraction	—	尚未评估
action: Write	0.7 fraction	—	尚未评估
action: Transfer	0.531 fraction	—	尚未评估

打开 Claim 详情

Action-conditioned trigger policies exhibit domain-dependent performance: on Retail with Qwen3-14B, triggering on Read (0.393) and Write (0.381) outperforms Respond (0.290), whereas on Airline with Qwen3-14B, the ordering inverts, with Respond achieving 0.372 while Read achieves 0.352 and Write achieves 0.340.整组结论：实验受阻，详见具体原因0/6 项指标获得支持 · 0 项已有评估6项测量结果展开结果表

The learned action classifier (predicting next action a_hat_t) is not released, nor are the training data, features, architecture, or tau-Bench integration harness.

测量结果

已评估 0/6 项

获得支持 0有争议或冲突 0尚未评估 6

保留同一结论下的完整结果组。已有评估排在前面；单项判断不替代整组结论。展开行查看条件、来源与历次证据。

每行是一项测量，分别显示论文报告、最近可用实测和累计证据状态。
指标与条件	论文报告	最近可用实测	证据判断
model: Qwen3-14B · domain: Retail · trigger_action: Read	0.393 fraction	—	尚未评估
model: Qwen3-14B · domain: Retail · trigger_action: Write	0.381 fraction	—	尚未评估
model: Qwen3-14B · domain: Retail · trigger_action: Respond	0.29 fraction	—	尚未评估
model: Qwen3-14B · domain: Airline · trigger_action: Respond	0.372 fraction	—	尚未评估
model: Qwen3-14B · domain: Airline · trigger_action: Read	0.352 fraction	—	尚未评估
model: Qwen3-14B · domain: Airline · trigger_action: Write	0.34 fraction	—	尚未评估

打开 Claim 详情

In the analysis of top-20 three-action chains, the information-acquisition chain Read->Read->Read (ID 1) accounts for 19.6%/22.2% (Correct/Incorrect, Rank #1) in Retail and 17.1%/14.5% (Rank #2) in Airline, while the repetitive output chain Resp->Resp->Resp (ID 2) accounts for 3.6%/8.2% (Rank #5) in Retail and 9.1%/23.6% (Rank #1) in Airline, showing strong error concentration in ungrounded conversational loops.整组结论：实验受阻，详见具体原因0/8 项指标获得支持 · 0 项已有评估8项测量结果展开结果表

Complete sliding window logs across episodes and domain trajectories are not available.

测量结果

已评估 0/8 项

获得支持 0有争议或冲突 0尚未评估 8

保留同一结论下的完整结果组。已有评估排在前面；单项判断不替代整组结论。展开行查看条件、来源与历次证据。

每行是一项测量，分别显示论文报告、最近可用实测和累计证据状态。
指标与条件	论文报告	最近可用实测	证据判断
group: correct · domain: Retail · pattern: Read->Read->Read · chain_id: 1	19.6 percentage_points	—	尚未评估
group: incorrect · domain: Retail · pattern: Read->Read->Read · chain_id: 1	22.2 percentage_points	—	尚未评估
group: correct · domain: Airline · pattern: Read->Read->Read · chain_id: 1	17.1 percentage_points	—	尚未评估
group: incorrect · domain: Airline · pattern: Read->Read->Read · chain_id: 1	14.5 percentage_points	—	尚未评估
group: correct · domain: Retail · pattern: Resp->Resp->Resp · chain_id: 2	3.6 percentage_points	—	尚未评估
group: incorrect · domain: Retail · pattern: Resp->Resp->Resp · chain_id: 2	8.2 percentage_points	—	尚未评估
group: correct · domain: Airline · pattern: Resp->Resp->Resp · chain_id: 2	9.1 percentage_points	—	尚未评估
group: incorrect · domain: Airline · pattern: Resp->Resp->Resp · chain_id: 2	23.6 percentage_points	—	尚未评估

打开 Claim 详情

Correct episodes exhibit higher variance slopes beta_t than incorrect episodes across all six task-LLM pairings. Separability is statistically significant (p < .001) for Math (Llama-3.1-8B), Code (Qwen3-14B), Retail (Qwen3-14B), and Retail (Qwen3-32B), but not statistically significant for Airline (Qwen3-14B, p=0.30; Qwen3-32B, p=0.057).整组结论：实验受阻，详见具体原因0/2 项指标获得支持 · 0 项已有评估2项测量结果展开结果表

Raw episode traces, prompt sharding scripts, and hidden state trajectory extraction codebase are unreleased.

测量结果

已评估 0/2 项

获得支持 0有争议或冲突 0尚未评估 2

保留同一结论下的完整结果组。已有评估排在前面；单项判断不替代整组结论。展开行查看条件、来源与历次证据。

每行是一项测量，分别显示论文报告、最近可用实测和累计证据状态。
指标与条件	论文报告	最近可用实测	证据判断
task: Airline · model: Qwen3-14B	0.3 scalar	—	尚未评估
task: Airline · model: Qwen3-32B	0.057 scalar	—	尚未评估

打开 Claim 详情

Across four domain-model settings on tau-Bench, geometry-conditioned adaptive reasoning triggers raise average task reward from 0.241 (24.1%) under Never-thinking to 0.396 (39.6%) while reducing mean token cost from 104.8k to 93.0k, an 11.2% reduction.整组结论：实验受阻，详见具体原因0/5 项指标获得支持 · 0 项已有评估5项测量结果展开结果表

The paper does not provide an official repository or executable code release (repository-identity.json indicates available: false). Furthermore, key protocol components including prompt templates, user simulator conversation driver scripts, probe extraction and logging pipelines, the exact classifier training split for Learned-HST, and detailed per-task configurations are deferred to an unreleased Appendix.

测量结果

已评估 0/5 项

获得支持 0有争议或冲突 0尚未评估 5

保留同一结论下的完整结果组。已有评估排在前面；单项判断不替代整组结论。展开行查看条件、来源与历次证据。

每行是一项测量，分别显示论文报告、最近可用实测和累计证据状态。
指标与条件	论文报告	最近可用实测	证据判断
policy: never-think · benchmark: tau-bench	0.241 fraction	—	尚未评估
policy: geometry-conditioned · benchmark: tau-bench	0.396 fraction	—	尚未评估
policy: never-think · benchmark: tau-bench	104.8 thousands_tokens	—	尚未评估
policy: geometry-conditioned · benchmark: tau-bench	93 thousands_tokens	—	尚未评估
benchmark: tau-bench	11.2 percentage_points	—	尚未评估

打开 Claim 详情

On tau-Bench Airline with Qwen3-32B, trigger policies yield task rewards of 0.136 (Never-think), 0.280 (Warm-up), 0.364 (Always-think), 0.328 (Variance slope beta_t > tau_beta), 0.352 (Temporal curvature kappa_t < -0.10), and 0.364 (Learned-HST), with mean token costs in thousands of 98.2, 92.6, 100.6, 86.3, 83.0, and 87.5 respectively.整组结论：实验受阻，详见具体原因0/18 项指标获得支持 · 0 项已有评估18项测量结果展开结果表

Official execution artifacts, dialogue simulators, user simulator policy prompts, and trajectory monitoring hooks are not released.

测量结果

已评估 0/18 项

获得支持 0有争议或冲突 0尚未评估 18

保留同一结论下的完整结果组。已有评估排在前面；单项判断不替代整组结论。展开行查看条件、来源与历次证据。

每行是一项测量，分别显示论文报告、最近可用实测和累计证据状态。
指标与条件	论文报告	最近可用实测	证据判断
model: Qwen3-32B · domain: Airline · policy: Never-think	0.136 fraction	—	尚未评估
model: Qwen3-32B · domain: Airline · policy: Warm-up	0.28 fraction	—	尚未评估
model: Qwen3-32B · domain: Airline · policy: Always-think	0.364 fraction	—	尚未评估
model: Qwen3-32B · domain: Airline · policy: Variance-slope	0.328 fraction	—	尚未评估
model: Qwen3-32B · domain: Airline · policy: Temporal-curvature · threshold: -0.1	0.352 fraction	—	尚未评估
model: Qwen3-32B · domain: Airline · policy: Learned-HST	0.364 fraction	—	尚未评估
model: Qwen3-32B · domain: Airline · policy: Never-think	98.2 thousands_tokens	—	尚未评估
model: Qwen3-32B · domain: Airline · policy: Warm-up	92.6 thousands_tokens	—	尚未评估
model: Qwen3-32B · domain: Airline · policy: Always-think	100.6 thousands_tokens	—	尚未评估
model: Qwen3-32B · domain: Airline · policy: Variance-slope	86.3 thousands_tokens	—	尚未评估
model: Qwen3-32B · domain: Airline · policy: Temporal-curvature · threshold: -0.1	83 thousands_tokens	—	尚未评估
model: Qwen3-32B · domain: Airline · policy: Learned-HST	87.5 thousands_tokens	—	尚未评估

打开 Claim 详情

On tau-Bench Airline with Qwen3-14B, trigger policies yield task rewards of 0.196 (Never-think), 0.348 (Warm-up), 0.356 (Always-think), 0.392 (Variance slope beta_t > tau_beta), 0.384 (Temporal curvature kappa_t < -0.15), and 0.420 (Learned-HST), with mean token costs in thousands of 82.5, 87.1, 83.5, 83.3, 78.9, and 85.0 respectively.整组结论：实验受阻，详见具体原因0/18 项指标获得支持 · 0 项已有评估18项测量结果展开结果表

Official execution artifacts, dialogue simulators, user simulator policy prompts, and trajectory monitoring hooks are not released.

测量结果

已评估 0/18 项

获得支持 0有争议或冲突 0尚未评估 18

保留同一结论下的完整结果组。已有评估排在前面；单项判断不替代整组结论。展开行查看条件、来源与历次证据。

每行是一项测量，分别显示论文报告、最近可用实测和累计证据状态。
指标与条件	论文报告	最近可用实测	证据判断
model: Qwen3-14B · domain: Airline · policy: Never-think	0.196 fraction	—	尚未评估
model: Qwen3-14B · domain: Airline · policy: Warm-up	0.348 fraction	—	尚未评估
model: Qwen3-14B · domain: Airline · policy: Always-think	0.356 fraction	—	尚未评估
model: Qwen3-14B · domain: Airline · policy: Variance-slope	0.392 fraction	—	尚未评估
model: Qwen3-14B · domain: Airline · policy: Temporal-curvature · threshold: -0.15	0.384 fraction	—	尚未评估
model: Qwen3-14B · domain: Airline · policy: Learned-HST	0.42 fraction	—	尚未评估
model: Qwen3-14B · domain: Airline · policy: Never-think	82.5 thousands_tokens	—	尚未评估
model: Qwen3-14B · domain: Airline · policy: Warm-up	87.1 thousands_tokens	—	尚未评估
model: Qwen3-14B · domain: Airline · policy: Always-think	83.5 thousands_tokens	—	尚未评估
model: Qwen3-14B · domain: Airline · policy: Variance-slope	83.3 thousands_tokens	—	尚未评估
model: Qwen3-14B · domain: Airline · policy: Temporal-curvature · threshold: -0.15	78.9 thousands_tokens	—	尚未评估
model: Qwen3-14B · domain: Airline · policy: Learned-HST	85 thousands_tokens	—	尚未评估

打开 Claim 详情

On tau-Bench Retail with Qwen3-32B, trigger policies yield task rewards of 0.347 (Never-think), 0.405 (Warm-up), 0.384 (Always-think), 0.412 (Variance slope beta_t > tau_beta), 0.442 (Temporal curvature kappa_t < -0.15), and 0.405 (Learned-HST), with mean token costs in thousands of 125.7, 107.1, 119.3, 105.1, 104.6, and 108.5 respectively.整组结论：实验受阻，详见具体原因0/18 项指标获得支持 · 0 项已有评估18项测量结果展开结果表

Official execution artifacts, dialogue simulators, user simulator policy prompts, and trajectory monitoring hooks are not released.

测量结果

已评估 0/18 项

获得支持 0有争议或冲突 0尚未评估 18

保留同一结论下的完整结果组。已有评估排在前面；单项判断不替代整组结论。展开行查看条件、来源与历次证据。

每行是一项测量，分别显示论文报告、最近可用实测和累计证据状态。
指标与条件	论文报告	最近可用实测	证据判断
model: Qwen3-32B · domain: Retail · policy: Never-think	0.347 fraction	—	尚未评估
model: Qwen3-32B · domain: Retail · policy: Warm-up	0.405 fraction	—	尚未评估
model: Qwen3-32B · domain: Retail · policy: Always-think	0.384 fraction	—	尚未评估
model: Qwen3-32B · domain: Retail · policy: Variance-slope	0.412 fraction	—	尚未评估
model: Qwen3-32B · domain: Retail · policy: Temporal-curvature · threshold: -0.15	0.442 fraction	—	尚未评估
model: Qwen3-32B · domain: Retail · policy: Learned-HST	0.405 fraction	—	尚未评估
model: Qwen3-32B · domain: Retail · policy: Never-think	125.7 thousands_tokens	—	尚未评估
model: Qwen3-32B · domain: Retail · policy: Warm-up	107.1 thousands_tokens	—	尚未评估
model: Qwen3-32B · domain: Retail · policy: Always-think	119.3 thousands_tokens	—	尚未评估
model: Qwen3-32B · domain: Retail · policy: Variance-slope	105.1 thousands_tokens	—	尚未评估
model: Qwen3-32B · domain: Retail · policy: Temporal-curvature · threshold: -0.15	104.6 thousands_tokens	—	尚未评估
model: Qwen3-32B · domain: Retail · policy: Learned-HST	108.5 thousands_tokens	—	尚未评估

打开 Claim 详情

Sweeping the temporal curvature threshold tau_kappa on Retail with Qwen3-14B yields task rewards of 0.388 at tau_kappa = -0.15, 0.397 at tau_kappa = -0.20, and 0.363 at tau_kappa = -0.25, demonstrating unimodal behavior.整组结论：实验受阻，详见具体原因0/3 项指标获得支持 · 0 项已有评估3项测量结果展开结果表

Evaluation scripts and environment setup for sweeping tau_kappa on tau-Bench Retail are not available.

测量结果

已评估 0/3 项

获得支持 0有争议或冲突 0尚未评估 3

保留同一结论下的完整结果组。已有评估排在前面；单项判断不替代整组结论。展开行查看条件、来源与历次证据。

每行是一项测量，分别显示论文报告、最近可用实测和累计证据状态。
指标与条件	论文报告	最近可用实测	证据判断
model: Qwen3-14B · domain: Retail · threshold: -0.15	0.388 fraction	—	尚未评估
model: Qwen3-14B · domain: Retail · threshold: -0.2	0.397 fraction	—	尚未评估
model: Qwen3-14B · domain: Retail · threshold: -0.25	0.363 fraction	—	尚未评估

打开 Claim 详情

In three-action chains, the proportions of chains containing each action differ between correct and incorrect episodes. In Retail, Write occurs in 49.3% vs 43.4% (Delta = -5.9%), Read in 85.8% vs 82.5% (Delta = -3.3%), Transfer in 4.2% vs 4.4% (Delta = +0.2%), and Respond in 57.4% vs 50.0% (Delta = -7.4%). In Airline, Write occurs in 15.3% vs 38.9% (Delta = +23.7%), Read in 86.1% vs 64.4% (Delta = -21.7%), Transfer in 29.2% vs 7.5% (Delta = -21.7%), and Respond in 58.3% vs 61.7% (Delta = +3.4%).整组结论：实验受阻，详见具体原因0/16 项指标获得支持 · 0 项已有评估16项测量结果展开结果表

The complete execution traces from which three-action sliding windows were extracted are not published, and action labeling mappings for tool calls are unspecified.

测量结果

已评估 0/16 项

获得支持 0有争议或冲突 0尚未评估 16

保留同一结论下的完整结果组。已有评估排在前面；单项判断不替代整组结论。展开行查看条件、来源与历次证据。

每行是一项测量，分别显示论文报告、最近可用实测和累计证据状态。
指标与条件	论文报告	最近可用实测	证据判断
group: correct · action: Write · domain: Retail	49.3 percentage_points	—	尚未评估
group: incorrect · action: Write · domain: Retail	43.4 percentage_points	—	尚未评估
group: correct · action: Read · domain: Retail	85.8 percentage_points	—	尚未评估
group: incorrect · action: Read · domain: Retail	82.5 percentage_points	—	尚未评估
group: correct · action: Transfer · domain: Retail	4.2 percentage_points	—	尚未评估
group: incorrect · action: Transfer · domain: Retail	4.4 percentage_points	—	尚未评估
group: correct · action: Respond · domain: Retail	57.4 percentage_points	—	尚未评估
group: incorrect · action: Respond · domain: Retail	50 percentage_points	—	尚未评估
group: correct · action: Write · domain: Airline	15.3 percentage_points	—	尚未评估
group: incorrect · action: Write · domain: Airline	38.9 percentage_points	—	尚未评估
group: correct · action: Read · domain: Airline	86.1 percentage_points	—	尚未评估
group: incorrect · action: Read · domain: Airline	64.4 percentage_points	—	尚未评估

打开 Claim 详情

On the Code task with Qwen3-14B, the correct group exhibits a positive median temporal curvature of +0.127 and a mean of +0.096, indicating that hidden-state updates on structured verifiable tasks tend toward same-direction accumulation rather than directional reversal.整组结论：实验受阻，详见具体原因0/2 项指标获得支持 · 0 项已有评估2项测量结果展开结果表

The Code task generation, interaction logs, and hidden-state extraction code are not available.

测量结果

已评估 0/2 项

获得支持 0有争议或冲突 0尚未评估 2

保留同一结论下的完整结果组。已有评估排在前面；单项判断不替代整组结论。展开行查看条件、来源与历次证据。

每行是一项测量，分别显示论文报告、最近可用实测和累计证据状态。
指标与条件	论文报告	最近可用实测	证据判断
task: Code · group: correct · model: Qwen3-14B	0.127 scalar	—	尚未评估
task: Code · group: correct · model: Qwen3-14B	0.096 scalar	—	尚未评估

打开 Claim 详情

In the turn-resolved hidden-state trajectory analysis on Math tasks (Figure 3), the sample sizes comprise n=259 correct episodes and n=227 incorrect episodes.整组结论：实验受阻，详见具体原因0/2 项指标获得支持 · 0 项已有评估2项测量结果展开结果表

The exact Math problem subset, seeds, sharding prompt splits, and model generation outputs from Lost in Conversation are not included in the paper snapshot.

测量结果

已评估 0/2 项

获得支持 0有争议或冲突 0尚未评估 2

保留同一结论下的完整结果组。已有评估排在前面；单项判断不替代整组结论。展开行查看条件、来源与历次证据。

每行是一项测量，分别显示论文报告、最近可用实测和累计证据状态。
指标与条件	论文报告	最近可用实测	证据判断
task: Math · group: correct	259 count	—	尚未评估
task: Math · group: incorrect	227 count	—	尚未评估

打开 Claim 详情

On tau-Bench Retail with Qwen3-14B, trigger policies yield task rewards of 0.286 (Never-think), 0.361 (Warm-up), 0.307 (Always-think), 0.393 (Variance slope beta_t > tau_beta), 0.397 (Temporal curvature kappa_t < -0.20), and 0.398 (Learned-HST), with mean token costs in thousands of 112.6, 111.1, 113.1, 104.3, 101.0, and 104.4 respectively.整组结论：实验受阻，详见具体原因0/18 项指标获得支持 · 0 项已有评估18项测量结果展开结果表

Execution requires official code and prompt/dialogue harnesses that are not published with the paper. Missing items include the tau-Bench agent loop modifications, the thinking mode activation protocol for Qwen3-14B, the layer 22 extraction hook, and the specific Learned-HST classifier model and weights.

测量结果

已评估 0/18 项

获得支持 0有争议或冲突 0尚未评估 18

保留同一结论下的完整结果组。已有评估排在前面；单项判断不替代整组结论。展开行查看条件、来源与历次证据。

每行是一项测量，分别显示论文报告、最近可用实测和累计证据状态。
指标与条件	论文报告	最近可用实测	证据判断
model: Qwen3-14B · domain: Retail · policy: Never-think	0.286 fraction	—	尚未评估
model: Qwen3-14B · domain: Retail · policy: Warm-up	0.361 fraction	—	尚未评估
model: Qwen3-14B · domain: Retail · policy: Always-think	0.307 fraction	—	尚未评估
model: Qwen3-14B · domain: Retail · policy: Variance-slope	0.393 fraction	—	尚未评估
model: Qwen3-14B · domain: Retail · policy: Temporal-curvature · threshold: -0.2	0.397 fraction	—	尚未评估
model: Qwen3-14B · domain: Retail · policy: Learned-HST	0.398 fraction	—	尚未评估
model: Qwen3-14B · domain: Retail · policy: Never-think	112.6 thousands_tokens	—	尚未评估
model: Qwen3-14B · domain: Retail · policy: Warm-up	111.1 thousands_tokens	—	尚未评估
model: Qwen3-14B · domain: Retail · policy: Always-think	113.1 thousands_tokens	—	尚未评估
model: Qwen3-14B · domain: Retail · policy: Variance-slope	104.3 thousands_tokens	—	尚未评估
model: Qwen3-14B · domain: Retail · policy: Temporal-curvature · threshold: -0.2	101 thousands_tokens	—	尚未评估
model: Qwen3-14B · domain: Retail · policy: Learned-HST	104.4 thousands_tokens	—	尚未评估

打开 Claim 详情

Geometry of Divergence: Tracking Hidden-State Trajectories for Adaptive Multi-Turn Reasoning

研究结论

复现结果

尚无复现结果

关键指标论文报告值 → 复现观测值 · 点击行展开详情

测量结果

测量结果

测量结果

测量结果

测量结果

测量结果

测量结果

测量结果

测量结果

测量结果

测量结果

测量结果

测量结果

测量结果