← 返回阅读笔记 / Back to reading note

A Human-Inspired Reading Agent with Gist Memory of Very Long Contexts (ReadAgent)

SUP-A01 原表

来源：作者原文。2026-09-07抓取，原始行列，不是统一复跑成绩。

Table 1: QuALITY results on the dev set of 230 docs and 2086 questions using PaLM 2-L. CR is the compression rate. # LU is the number of lookups. We report means and standard deviations across 3 runs, except where inconsequential (CR and # LU).
Method CR (# LU) Accuracy
BM25 Retrieval
      Top-1 89.27% (1) 70.34% ±\pm 0.06
      Top-2 78.96% (2) 79.05% ±\pm 0.05
      Top-3 68.50% (3) 82.65% ±\pm 0.05
      Top-4 58.57% (4) 84.42% ±\pm 0.13
Neural Retrieval with Gemini API
      Top-1 89.91% (1) 71.32% ±\pm 0.19
      Top-2 80.08% (2) 79.02% ±\pm 0.10
      Top-3 70.28% (3) 83.41% ±\pm 0.10
      Top-4 60.68% (4) 84.88% ±\pm 0.03
Full Raw Content 0% 85.83% ±\pm 0.19
GistMem 85.53% 77.52% ±\pm 0.13
ReadAgent-P
      Look up 1 pg 76.00% (1.0) 84.13% ±\pm 0.10
      Look up 1-2 pgs 72.17% (1.6) 86.16% ±\pm 0.12
      Look up 1-3 pgs 69.36% (2.0) 86.59% ±\pm 0.10
      Look up 1-4 pgs 67.73% (2.2) 86.86% ±\pm 0.00
      Look up 1-5 pgs 66.45% (2.3) 86.83% ±\pm 0.10
      Look up 1-6 pgs 64.75% (2.5) 86.91% ±\pm 0.08
ReadAgent-S 1-6 pgs 58.53% (3.2) 87.17% ±\pm 0.18
Table 2: NarrativeQA results (PaLM 2-L). R-1 , R-2 , and R-L are ROUGE F-Measures. LR-1 , and LR-2 are LLM-Ratings.
Gutenberg Validation (58 docs & 1743 questions) Gutenberg Test (177 docs & 5207 questions)
Method CR (# LU) LR-1 LR-2 R-1 R-2 R-L CR (# LU) LR-1 LR-2 R-1 R-2 R-L
BM25 Retrieval
Top-1 97.63% (1) 39.01% 50.14% 0.166 0.061 0.156 97.42% (1) 43.5% 55.33% 0.176 0.065 0.165
Top-2 95.24% (2) 49.34% 60.76% 0.203 0.079 0.191 94.80% (2) 51.70% 64.53% 0.206 0.082 0.194
Top-3 93.34% (3) 52.73% 63.68% 0.208 0.080 0.195 93.02% (3) 52.97% 66.03% 0.210 0.083 0.197
Top-4 92.47% (4) 53.59% 64.26% 0.211 0.082 0.197 92.27% (4) 53.60% 66.16% 0.210 0.084 0.197
Neural Retrieval with Gemini API
Top-1 98.19% (1) 34.25% 46.53% 0.146 0.051 0.134 98.14% (1) 36.47% 47.8% 0.150 0.054 0.140
Top-2 96.30% (2) 44.69% 54.96% 0.180 0.069 0.167 96.15% (2) 44.48% 56.17% 0.182 0.070 0.170
Top-3 94.62% (3) 46.24% 57.31% 0.191 0.077 0.178 94.42% (3) 48.97% 60.73% 0.195 0.076 0.183
Top-4 93.45% (4) 48.59% 59.21% 0.196 0.079 0.184 93.25% (4) 50.62% 62.05% 0.203 0.080 0.191
GistMem 96.89% 55.31% 68.22% 0.233 0.091 0.218 96.80% 55.79% 71.19% 0.231 0.092 0.217
ReadAgent-P
Look up 1 pg 95.15% (0.94) 58.92% 71.89% 0.244 0.101 0.230 94.84% (0.93) 59.98% 73.23% 0.240 0.098 0.226
Look up 1-2 pgs 94.79% (1.23) 59.84% 72.29% 0.239 0.098 0.224 94.36% (1.34) 59.19% 72.65% 0.231 0.091 0.218
Look up 1-3 pgs 94.39% (1.50) 59.84% 71.89% 0.240 0.098 0.226 94.03% (1.61) 59.63% 72.84% 0.230 0.093 0.217
ReadAgent-S 1-2 pgs 94.35% (1.38) 57.89% 71.14% 0.239 0.097 0.225 93.86% (1.46) 60.48% 72.48% 0.232 0.095 0.219
ReadAgent-S 1-3 pgs 94.08% (1.57) 58.52% 71.49% 0.242 0.098 0.229 93.67% (1.57) 60.55% 72.79% 0.231 0.095 0.219
Movie Validation (57 docs & 1699 questions) Movie Test (172 docs & 5139 questions)
BM25 Retrieval
Top-1 97.07% (1) 32.67% 42.61% 0.156 0.058 0.144 96.61% (1) 33.64% 43.34% 0.154 0.054 0.143
Top-2 94.12% (2) 39.97% 50.21% 0.187 0.070 0.174 93.81% (2) 42.50% 53.05% 0.191 0.072 0.178
Top-3 91.18% (3) 43.61% 53.91% 0.198 0.077 0.185 91.00% (3) 46.97% 57.52% 0.207 0.080 0.193
Top-4 88.24% (4) 46.85% 57.62% 0.210 0.084 0.198 88.19% (4) 50.18% 60.13% 0.217 0.085 0.202
Neural Retrieval with Gemini API
Top-1 97.07% (1) 32.02% 41.44% 0.153 0.053 0.142 96.67% (1) 37.24% 46.22% 0.130 0.043 0.118
Top-2 94.19% (2) 43.20% 51.38% 0.160 0.057 0.148 93.90% (2) 46.49% 54.60% 0.164 0.061 0.151
Top-3 91.29% (3) 47.56% 56.21% 0.176 0.064 0.163 91.14% (3) 50.69% 58.92% 0.186 0.071 0.172
Top-4 88.38% (4) 49.09% 59.33% 0.193 0.075 0.180 88.36% (4) 52.13% 59.41% 0.184 0.072 0.171
GistMem 92.09% 52.56% 64.39% 0.242 0.103 0.227 91.98% 54.68% 64.00% 0.248 0.105 0.234
ReadAgent-P
Look up 1 pg 89.20% (0.99) 53.38% 65.57% 0.247 0.106 0.233 89.22% (0.98) 57.68% 68.01% 0.274 0.116 0.260
Look up 1-2 pgs 87.68% (1.52) 54.62% 65.63% 0.238 0.098 0.223 88.10% (1.39) 58.24% 68.81% 0.270 0.115 0.255
Look up 1-3 pgs 86.57% (1.91) 54.91% 65.86% 0.241 0.099 0.225 86.73% (1.89) 58.82% 69.12% 0.272 0.116 0.257
ReadAgent-S 1-2 pgs 86.36% (1.98) 59.33% 68.28% 0.203 0.082 0.188 85.92% (1.98) 63.33% 72.06% 0.214 0.086 0.199
ReadAgent-S 1-3 pgs 83.56% (2.95) 59.45% 68.81% 0.210 0.087 0.195 83.18% (2.95) 64.53% 73.06% 0.217 0.090 0.202
Table 3: QMSum validation results (PaLM 2-L) means and standard deviations across 3 runs. 35 articles and 272 questions. CR is the compression rate. # LU is the number of lookups. Resp. Length is the length in words of the model’s final response.
Method CR (# LU) LLM Rating-1 LLM Rating-2 ROUGE-1 ROUGE-2 ROUGE-L Resp. Length
BM25 Retrieval
Top-1 95.69% (1.00) 32.48% ±\pm 1.65 63.85% ±\pm 1.51 27.53 ±\pm 0.23 7.00 ±\pm 0.14 18.45 ±\pm 0.16 48.62 ±\pm 0.28
Top-2 91.48% (2.00) 29.41% ±\pm 0.60 71.57% ±\pm 1.48 28.85 ±\pm 0.17 7.59 ±\pm 0.08 19.34 ±\pm 0.14 52.39 ±\pm 0.49
Top-3 86.93% (3.00) 34.80% ±\pm 1.14 79.53% ±\pm 0.35 30.69 ±\pm 0.17 8.40 ±\pm 0.11 20.64 ±\pm 0.13 53.59 ±\pm 0.35
Top-4 82.55% (4.00) 35.66% ±\pm 0.30 81.13% ±\pm 0.35 31.10 ±\pm 0.10 8.53 ±\pm 0.06 20.36 ±\pm 0.11 54.96 ±\pm 0.42
Top-5 78.13% (5.00) 39.09% ±\pm 0.92 84.44% ±\pm 0.46 31.16 ±\pm 0.14 8.52 ±\pm 0.08 20.69 ±\pm 0.03 54.52 ±\pm 0.13
Top-6 73.97% (6.00) 37.87% ±\pm 0.90 83.70% ±\pm 0.87 31.06 ±\pm 0.04 8.38 ±\pm 0.06 20.43 ±\pm 0.08 56.18 ±\pm 0.44
Neural Retrieval with Gemini API
Top-1 95.99% (1.00) 34.80% ±\pm 1.39 68.87% ±\pm 0.62 27.86 ±\pm 0.12 7.12 ±\pm 0.04 18.76 ±\pm 0.09 49.46 ±\pm 0.23
Top-2 92.02% (2.00) 40.32% ±\pm 0.92 81.50% ±\pm 0.46 30.17 ±\pm 0.08 8.03 ±\pm 0.03 19.80 ±\pm 0.08 55.48 ±\pm 0.27
Top-3 87.93% (3.00) 40.93% ±\pm 1.35 85.17% ±\pm 1.25 31.36 ±\pm 0.12 8.67 ±\pm 0.10 20.68 ±\pm 0.10 56.71 ±\pm 0.27
Top-4 83.71% (4.00) 40.56% ±\pm 0.62 84.31% ±\pm 0.87 31.52 ±\pm 0.11 8.59 ±\pm 0.10 20.40 ±\pm 0.10 56.47 ±\pm 0.71
Top-5 79.47% (5.00) 40.20% ±\pm 0.76 86.76% ±\pm 0.60 31.32 ±\pm 0.11 8.49 ±\pm 0.11 20.49 ±\pm 0.07 56.73 ±\pm 0.91
Top-6 75.44% (6.00) 40.81% ±\pm 0.52 87.01% ±\pm 0.35 31.92 ±\pm 0.02 8.73 ±\pm 0.09 20.82 ±\pm 0.05 58.39 ±\pm 0.31
Truncated Raw Content
First 6k words 32.59% (0.00) 14.71% ±\pm 0.79 52.45% ±\pm 0.69 25.42 ±\pm 0.05 4.98 ±\pm 0.09 16.58 ±\pm 0.10 58.42 ±\pm 0.11
Last 6k words 32.38% (0.00) 10.42% ±\pm 0.62 35.66% ±\pm 2.46 20.69 ±\pm 0.19 3.44 ±\pm 0.10 14.13 ±\pm 0.08 44.23 ±\pm 0.11
GistMem 83.13% (0.00) 40.20% ±\pm 0.96 89.83% ±\pm 0.76 31.00 ±\pm 0.09 7.99 ±\pm 0.04 20.15 ±\pm 0.08 65.75 ±\pm 0.20
ReadAgent-P
Look up 1 pg 80.00% (0.98) 40.56% ±\pm 0.46 89.46% ±\pm 1.48 31.26 ±\pm 0.09 8.22 ±\pm 0.15 20.29 ±\pm 0.07 63.78 ±\pm 1.13
Look up 1-2 pgs 77.38% (1.71) 39.71% ±\pm 1.87 89.71% ±\pm 0.60 31.11 ±\pm 0.04 8.01 ±\pm 0.15 20.21 ±\pm 0.04 64.73 ±\pm 1.02
Look up 1-3 pgs 75.07% (2.53) 38.36% ±\pm 1.21 89.71% ±\pm 0.60 31.50 ±\pm 0.29 8.15 ±\pm 0.15 20.45 ±\pm 0.24 63.91 ±\pm 1.58
Look up 1-4 pgs 73.48% (3.08) 39.95% ±\pm 1.51 90.56% ±\pm 0.35 31.34 ±\pm 0.05 8.08 ±\pm 0.18 20.26 ±\pm 0.07 63.40 ±\pm 0.79
Look up 1-5 pgs 72.29% (3.50) 37.99% ±\pm 0.96 87.75% ±\pm 0.46 31.16 ±\pm 0.10 8.06 ±\pm 0.05 20.35 ±\pm 0.12 65.22 ±\pm 1.40
Look up 1-6 pgs 70.90% (3.97) 39.09% ±\pm 2.04 88.24% ±\pm 0.60 31.50 ±\pm 0.30 8.05 ±\pm 0.13 20.26 ±\pm 0.13 66.70 ±\pm 0.62
ReadAgent-S 1-6 pgs 70.34% (3.55) 46.57% ±\pm 0.87 91.54% ±\pm 0.30 32.90 ±\pm 0.17 8.87 ±\pm 0.23 21.15 ±\pm 0.14 68.87 ±\pm 0.60
Table 4: ReadAgent retrieval vs. GistMem with neural retrieval.
Method Accuracy
GistMem + Neural Retrieval Top-1 82.65%
ReadAgent-P (Look up 1 pg) 84.13%
Table 5: ReadAgent accuracy on QuALITY with episode pagination based on LLM (PaLM 2-L) vs. uniform length pagination.
LLM Uniform Length
ReadAgent-P (1-5 pgs) Acc. 86.83% 85.71%
Table 6: Compression rate increases as the maximum number of words allowed per page increases on QuALITY. Our default setting of min/max words is 280/600. In the other three experiments, we scale min words proportionally with max words.
GistMem ReadAgent-P (1-5 pgs)
max_words CR Acc CR Acc
400 81.81% 78.91% 66.71% 86.82%
600 85.53% 77.52% 66.45% 86.83%
800 88.12% 76.22% 65.06% 86.34%
1200 91.38% 73.97% 61.77% 85.67%
Table 7: QuALITY results on the dev set of 230 docs and 2086 questions using GPT-3.5-turbo. CR is the compression rate. # LU is the number of lookups. We report 1 run for each experiment for cost considerations.
Method CR (# LU) Accuracy
Neural Retrieval with Gemini API Top-3 73.13% (3) 69.22%
Full Raw Content 0% 73.30%
GistMem 84.24% 66.06%
ReadAgent-P Look up 1-5 pgs 76.60% (1.0) 69.65%
ReadAgent-S Look up 1-6 pgs 60.43% (3.4) 72.10%
Table 8: Pagination hyperparameters.
Dataset max_wordsmin_words\frac{\text{{max\textunderscore words}}}{\text{{min\textunderscore words}}}
QuALITY 600280\frac{600}{280}
QMSum 600280\frac{600}{280}
NarrativeQA Gutenberg 3000500\frac{3000}{500}
NarrativeQA movie scripts 1000600\frac{1000}{600}
Table 9: Pagination for “off course”.
Page # Starting sentence in text
0 Shure and begorra…
1 The alien stooped down…
2 Interest in the horse was ended…
3 “Sure, and it’s quite a reception”…
4 Excitement, anticipation…
5 “Here he comes,”…
6 He straightened and started off…
7 The others drew back…
Table 10: Web navigation performance on Mind2Web ( Deng et al., 2023 ) . ∗ marks models that are trained supervisedly for the web domain. GistMem and ReadAgent results are all also based on PaLM 2-L. We evaluate the performance in element accuracy (Ele. Acc), operation F1 (Op. F1), step success rate (Step SR), and episode success rate (SR). We also measure the compression rate (CR). The best performance across all the baselines is bolded , and the best across the approaches using PaLM 2-L is underlined . ReadAgent achieves consistently better performance than using raw HTML inputs (PaLM 2-L), retrieval methods, and MindAct (PaLM 2-L) with a trained Rank LM for HTML snippet retrieval.
Cross-Task Cross-Website Cross-Domain
CR Ele. Acc Op. F1 Step SR SR CR Ele. Acc Op. F1 Step SR SR CR Ele. Acc Op. F1 Step SR SR
Using supervisedly trained RankLM
MindAct (GPT-3.5 + Rank LM∗) – 20.3 56.6 17.4 0.8 – 19.3 48.8 16.2 0.6 – 21.6 52.8 18.6 1.0
MindAct (GPT-4 + Rank LM∗) – 41.6 60.6 36.2 2.0 – 35.8 51.1 30.1 2.0 – 37.1 46.5 26.4 2.0
MindAct (PaLM 2-L + Rank LM∗) – 29.8 61.9 24.4 1.2 – 28.8 59.6 21.6 0.6 – 29.9 60.4 24.5 1.3
No training
(PaLM 2-L)
+Raw HTML 0.0 22.1 76.7 19.2 1.2 0.0 22.2 72.3 18.2 1.7 0.0 23.6 75.6 20.9 1.0
+BM25 Retrieval (Top-1) 43.7 16.3 61.7 14.2 0.4 49.7 17.8 60.8 15.2 0.0 51.6 17.3 60.4 15.9 0.0
+BM25 Retrieval (Top-5) 19.5 25.9 70.4 22.4 2.0 17.6 29.5 71.8 23.1 1.7 19.2 27.6 71.1 24.4 1.0
+Neural Retrieval (Top-1) 74.4 14.6 55.5 11.7 0.4 87.9 18.0 55.8 14.0 0.0 82.8 16.4 60.3 14.2 0.1
+Neural Retrieval (Top-5) 32.4 26.4 71.9 22.6 0.8 37.2 26.7 69.1 22.3 2.8 38.1 30.0 72.5 26.9 1.2
GistMem 84.4 11.7 43.1 9.5 0.0 82.5 11.7 43.6 8.4 0.0 83.0 13.4 49.6 11.7 0.5
ReadAgent-P: Lookup 1 snippet 55.1 31.1 70.1 26.8 2.0 54.1 34.5 74.1 28.2 2.3 55.2 36.1 75.6 33.0 2.0
ReadAgent-P: Lookup 1-5 snippets 35.9 33.7 72.5 29.2 2.8 35.6 37.4 75.1 31.1 3.4 48.2 37.2 76.3 33.4 2.3
Δ⁡(Raw−ReadAgent)\Delta(\text{Raw}-\text{ReadAgent}) – +11.6 -4.2 +10.0 +1.6 – +15.2 +2.8 +12.9 +1.7 – +13.6 +0.7 +12.5 +1.3
Δ⁡(MindAct−ReadAgent)\Delta(\text{MindAct}-\text{ReadAgent}) – +3.9 +10.6 +4.8 +1.6 – +8.6 +15.5 +9.5 +2.8 – +7.3 +15.9 +8.9 +1.0
Figure 6: (Left) Histogram of raw HTML and gist tokens in the Mind2Web cross-website split. Most of the input gists require fewer than 8K tokens. (Right) Statistics of token counts of raw HTML and gists.
Threshold Raw Gist
4096 Tokens 14.2% 88.6%
8192 Tokens 51.5% 97.4%
16384 Tokens 79.1% 100%
50th Percentile Tokens 8018 989
90th Percentile Tokens 25337 3596
95th Percentile Tokens 35779 5741
99th Percentile Tokens 55642 12569
Table 11: QMSum test results (PaLM 2-L) means and standard deviations across 3 runs. 35 articles and 281 questions. Bold methods are this work. Bold values are the best; bold italics are ties for best. CR is the compression rate. # LU is the number of lookups. Resp. Length is the length in words of the model’s final response. We omit standard deviations for CR and # LU for presentation purposes; they were all inconsequential.
Method CR (# LU) LLM Rating-1 LLM Rating-2 ROUGE-1 ROUGE-2 ROUGE-L Resp. Length
BM25 Retrieval
Top-1 95.61% (1.00) 24.67% ±\pm 0.44 66.90% ±\pm 0.87 28.81 ±\pm 0.13 8.14 ±\pm 0.15 19.62 ±\pm 0.18 48.15 ±\pm 0.18
Top-2 91.32% (2.00) 31.79% ±\pm 1.31 79.95% ±\pm 0.67 30.89 ±\pm 0.13 9.14 ±\pm 0.05 20.67 ±\pm 0.09 53.91 ±\pm 0.64
Top-3 87.25% (3.00) 33.45% ±\pm 0.00 83.63% ±\pm 1.05 31.39 ±\pm 0.23 9.11 ±\pm 0.05 21.03 ±\pm 0.03 55.15 ±\pm 0.61
Top-4 82.86% (4.00) 37.72% ±\pm 1.05 86.12% ±\pm 0.50 31.71 ±\pm 0.09 9.35 ±\pm 0.13 21.26 ±\pm 0.12 58.21 ±\pm 0.37
Top-5 78.79% (5.00) 39.38% ±\pm 1.02 86.60% ±\pm 0.44 32.66 ±\pm 0.04 9.98 ±\pm 0.10 21.86 ±\pm 0.05 59.20 ±\pm 1.05
Top-6 74.62% (6.00) 40.45% ±\pm 0.89 90.98% ±\pm 0.34 32.56 ±\pm 0.03 9.78 ±\pm 0.03 21.64 ±\pm 0.09 60.40 ±\pm 1.28
Neural Retrieval with Gemini API
Top-1 95.80% (1.00) 27.05% ±\pm 0.50 67.97% ±\pm 1.74 28.71 ±\pm 0.12 7.98 ±\pm 0.04 19.59 ±\pm 0.04 49.76 ±\pm 0.78
Top-2 91.62% (2.00) 35.35% ±\pm 0.44 80.07% ±\pm 0.00 31.65 ±\pm 0.18 9.59 ±\pm 0.11 21.29 ±\pm 0.11 56.19 ±\pm 0.76
Top-3 87.39% (3.00) 35.71% ±\pm 1.37 88.49% ±\pm 0.34 32.33 ±\pm 0.17 9.84 ±\pm 0.07 21.54 ±\pm 0.13 59.19 ±\pm 0.96
Top-4 83.28% (4.00) 39.62% ±\pm 0.17 90.15% ±\pm 0.34 32.31 ±\pm 0.21 9.69 ±\pm 0.15 21.65 ±\pm 0.15 59.86 ±\pm 0.11
Top-5 79.33% (5.00) 44.01% ±\pm 0.84 91.22% ±\pm 0.34 32.33 ±\pm 0.24 9.84 ±\pm 0.21 21.67 ±\pm 0.19 61.53 ±\pm 0.35
Top-6 75.35% (6.00) 44.60% ±\pm 0.89 92.65% ±\pm 0.17 32.55 ±\pm 0.08 9.75 ±\pm 0.21 21.39 ±\pm 0.13 61.29 ±\pm 0.46
Truncated Raw Content
First 6k words 31.51% (0.00) 13.17% ±\pm 1.05 47.81% ±\pm 5.90 24.15 ±\pm 1.42 4.89 ±\pm 0.57 16.27 ±\pm 0.96 61.43 ±\pm 3.53
Last 6k words 33.80% (0.00) 13.76% ±\pm 0.84 43.42% ±\pm 0.00 22.90 ±\pm 0.10 4.35 ±\pm 0.04 15.69 ±\pm 0.03 52.47 ±\pm 0.39
GistMem 82.81% (0.00) 44.96% ±\pm 0.44 91.93% ±\pm 0.73 31.20 ±\pm 0.17 9.02 ±\pm 0.09 20.60 ±\pm 0.14 65.84 ±\pm 0.87
ReadAgent-P
Look up 1 pg 79.37% (0.98) 44.84% ±\pm 0.00 92.29% ±\pm 0.34 31.46 ±\pm 0.12 9.09 ±\pm 0.11 20.63 ±\pm 0.05 66.74 ±\pm 0.74
Look up 1-2 pgs 77.00% (1.72) 43.42% ±\pm 1.01 92.88% ±\pm 1.05 31.77 ±\pm 0.16 9.11 ±\pm 0.12 20.70 ±\pm 0.08 65.55 ±\pm 0.28
Look up 1-3 pgs 74.85% (2.46) 44.37% ±\pm 1.21 91.22% ±\pm 0.44 31.89 ±\pm 0.06 8.98 ±\pm 0.13 20.70 ±\pm 0.09 66.06 ±\pm 1.63
Look up 1-4 pgs 73.26% (3.02) 44.13% ±\pm 0.50 90.51% ±\pm 0.44 31.87 ±\pm 0.07 9.12 ±\pm 0.06 20.77 ±\pm 0.01 66.44 ±\pm 0.74
Look up 1-5 pgs 72.01% (3.44) 43.42% ±\pm 1.45 91.22% ±\pm 0.60 31.80 ±\pm 0.16 9.03 ±\pm 0.07 20.64 ±\pm 0.03 66.48 ±\pm 0.39
Look up 1-6 pgs 70.65% (3.89) 42.70% ±\pm 1.54 90.51% ±\pm 0.73 31.74 ±\pm 0.09 8.90 ±\pm 0.09 20.66 ±\pm 0.16 66.24 ±\pm 1.14
ReadAgent-S 1-6 pgs 70.75% (3.42) 49.58% ±\pm 0.44 93.83% ±\pm 0.34 32.88 ±\pm 0.15 9.98 ±\pm 0.06 21.50 ±\pm 0.04 67.86 ±\pm 0.11